key: cord-0841942-cssgzvus authors: Sivakumaran, K; Kim, Chul-Hyun; Tayon, Robert; Kao, C.Cheng title: RNA sequence and secondary structural determinants in a minimal viral promoter that directs replicase recognition and initiation of genomic plus-strand RNA synthesis 1 1 Edited by D. E. Draper date: 1999-12-03 journal: Journal of Molecular Biology DOI: 10.1006/jmbi.1999.3297 sha: 8a835eec68c335d412ed7af32a557373771bfd59 doc_id: 841942 cord_uid: cssgzvus Abstract Viral RNA replication provides a useful system to study the structure and function of RNAs and the mechanism of RNA synthesis from RNA templates. Previously we demonstrated that a 27 nt RNA from brome mosaic virus (BMV) can direct correct initiation of genomic plus-strand RNA synthesis by the BMV replicase. In this study, using biochemical, nuclear magnetic resonance, and thermodynamic analyses, we determined that the secondary structure of this 27 nt RNA can be significantly altered and retain the ability to direct RNA synthesis. In contrast, we find that position-specific changes in the RNA sequence will affect replicase recognition, modulate the polymerization process, and contribute to the differential accumulation of viral RNAs. These functional results are in agreement with the phylogenetic analysis of BMV and related viral sequences and suggest that a similar mechanism of RNA synthesis takes place for members of the alphavirus superfamily. RNA-dependent RNA polymerases (RdRps) are an interesting class of enzymes for several reasons: (1) RdRps are key enzymes in the replication of viral pathogens, including the causative agents of measles, the common cold, and viral hepatitis (Van Dyke & Flanegan, 1980; Neulfeld et al., 1991; Poch et al., 1990; Behrens et al., 1996; Yuan et al., 1997) ; (2) cellular RdRps affect gene expression in enucleated cells and in gene silencing (Volloch et al., 1996; Bingham, 1997; Schiebel et al., 1998; Cogoni & Macino, 1999) ; and (3) RdRps are the only class of template-dependent polymerases that can initiate synthesis de novo from the end of the template (for reviews, see Buck, 1996; O'Reilly & Kao, 1998) . We study the mechanism of RNA synthesis and promoter recognition using virally encoded replicases. The virally encoded RdRp, along with viral and cellular proteins make up the replicase of positivesense RNA viruses (Hardy et al., 1979; Quadt & Jaspars, 1990; Buck, 1996) . In one cycle of RNA replication, the replicase uses the genomic plusstrand RNA as the template to generate the complementary minus-strand RNA, which in turn serves as the template for generating multiple copies of genomic, and in some viral species, subgenomic plus-strand RNAs. Viral RNA replication requires the speci®c recognition of RNA features by the replicase, a process that is only beginning to be elucidated. We study RNA synthesis using the model system brome mosaic virus (BMV), a plant-infecting member of the alphavirus-like superfamily of positive-strand RNA viruses (Goldbach et al., 1991) . BMV has three genomic RNAs, designated as RNA1 (3.2 kb), RNA2 (2.8 kb), and RNA3 (2.1 kb). RNA1 and RNA2 are monocistronic and encode the 1a and 2a proteins, respectively, which are essential for the replication of the BMV genome (Ahlquist, 1992) . RNA3 is dicistronic and encodes the 3a movement protein and the coat protein that is expressed via a subgenomic RNA. An in vitro system using highly enriched BMV replicase preparations from infected barley has been developed to study initiation of RNA synthesis from minimal templates. Using this system, accurate initiation of minus-strand, genomic plus-strand and subgenomic RNAs in vitro has been demonstrated (Dreher & Hall, 1988; Miller et al., 1985; Sun et al., 1996; Adkins et al., 1997; Siegel et al., 1997 Siegel et al., , 1998 Chapman et al., 1998; Sivakumaran & Kao, 1999) . The initiation of genomic plus-strand RNA synthesis from the 3 H end of the minus-strand template requires one nucleotide 3 H of the initiation nucleotide (Sivakumaran & Kao, 1999) . This minus-strand template is recognized by the BMV replicase in a species-speci®c manner; the BMV replicase does not ef®ciently recognize the (À)-strand RNA of the closely related cucumber mosaic virus (CMV) or tobacco mosaic virus (Sivakumaran & Kao, 1999; Chapman & Kao, 1999) . In this paper, the secondary structure and sequence determinants governing the speci®c recognition of the RNA directing genomic plus-strand RNA synthesis was investigated. Using biochemical analysis and nuclear magnetic resonance spectroscopy (NMR), we found that ef®cient RNA-directed genomic plus-strand RNA synthesis can take place with more than one RNA secondary structure. Furthermore, the template sequence appears to determine the ef®ciency of RNA synthesis. Proposed secondary structure required to direct genomic plus-strand synthesis We have demonstrated that accurate initiation of genomic plus-strand synthesis from input minusstrand templates of 27 nt was comparable to those from longer templates (Sivakumaran & Kao, 1999) . The removal of two additional 5 H template nucleotides from the 27 nt RNA reduced synthesis to 30 %. Therefore, the 27 nt RNA B2(À)26G represents a minimal functional promoter and template (to be named an endscript since the promoter is also the template). We seek to use endscript B2(À)26G to determine the features that enable ef®cient plus-strand RNA synthesis. Multidimensional NMR is a powerful method for detailed structural analysis of molecules in solution. By virtue of its small size (27 nt), B2(À)26G is amenable to studies with proton NMR. NMR analysis requires the RNA to be in one predominant conformation. Since the regions forming intramolecular base-pairing can have high potential to form a dimeric conformation through intermolecular base-pairing, it is important to ensure that the RNA forms a stable monomeric conformation. To examine this, UV melting experiments were done with B2(À)26G over a 30-fold concentration range. Melting temperature (t m ) for the major transition was estimated at 55.6(AE0.5) C, and did not change appreciably over a concentration ranging from 2 to 60 mM, indicating that B2(À)26G has a monomeric conformation. In addition, native gel analysis was performed to determine the number of electrophoretically distinct conformations of B2(À)26G that may exist. As a control, a 17 nt RNA that exists both as a monomer and a dimer was used (Figure 1(a) , lane 1). B2(À)26G was found to migrate predominantly as a single band (Figure 1(a) , lane 2), indicating that it does not form a dimer. Also, when tested in salt concentrations ranging from 50 to 200 mM, the mobility did not change, indicating that its conformation is stable in the range of salt conditions used for NMR studies. The most stable secondary structure predicted for B2(À)26G by the Zuker mfold program (Jaeger et al., 1989) is shown in Figure 1 (b). We sought to con®rm or modify this predicted structure using the results obtained from NMR analysis. Although solving a complete tertiary structure at atomic resolution using NMR is a time-consuming process, one-dimensional (1D) imino NMR and two dimensional NMR spectra of an RNA molecule can be used to determine RNA secondary structure (Varani & Tinoco, 1991) . When the imino protons are involved in base-pairing or forming an unusual conformation, they are usually protected from fast proton exchange with the water solvent, resulting in sharp imino proton peaks. If those protected imino protons are located within a 5 A Ê space, the crosspeaks between imino peaks can be observed by two-dimensional nuclear Overhauser effect spectroscopy (2D NOESY), where protected imino protons in guanine and uracil bases will result in a NOE in a NOESY spectrum (Variani & Tinoco, 1991) and allow the assignment of nucleotides in base-pairs. As seen in Figure 1 (b), the proposed secondary structure contains eight base-pairs with a UÁ U mismatch and a CUU tri-loop. If this structure is the major conformation in solution, we should be able to see major imino peaks from the two A Á U, four GÁ C, and one GÁ U base-pair. The terminal A Á U base-pair might not be stable due to increased¯exibility caused by the adjacent 5 H and 3 H overhangs. More than eight imino peaks are observed in the 1D imino spectrum (Figure 1(c) ), which indicates that the 27 nt RNA may have unusually protected imino peaks or some minor conformation that can be clari®ed by analyzing 2D water NOESY crosspeak pattern. The crosspeak pattern shown in the 2D H 2 O NOESY (Figure 1(d) ) would allow us to discern between GÁ C and A Á U base-pairs using the characteristic crosspeaks between the uracil imino and adenine H2 in the A Á U base-pair, and the crosspeaks between guanine imino and cytosine amino protons in the G ÁC base-pair, respectively. The expected GÁ U wobble base-pair was discerned by its characteristic strong crosspeaks between G and U imino protons. The connectivity between weaker AÁU base-pairs was also observed (gray circles with dotted lines in Figure 1 (d)), which might be either the minor conformation of A Á U base-pairs or unstable A Á U base-pairs from the terminal stem or 5 H overhang. Interestingly, the U Á U mismatched base-pairs were observed and con®rmed by its characteristic strong crosspeaks between the N3 proton of base-paired uracil bases and its connectivity with guanine 7 imino proton. Overall, with the possible presence of its local minor conformation, the assignment of imino peaks based on 1D and 2D NMR is well matched with the predicted secondary structure of B2(À)26G (Figure 1(b) ). Effect of mutations on synthesis from B2(À)26G endscript Mutational analysis of B2(À)26G was performed to determine the features required for recognition by the BMV replicase. Mutant endscripts were all quanti®ed after transcription and puri®cation as described by . In essence, one picomole of each endscript was subjected to RNA synthesis assay and the amount of product synthesized was normalized to that produced by wildtype B2(À)26G. B2(À)26G yielded a predominant product of 26 nt, and some minor amounts of RNA one nucleotide longer and shorter due to non-templated nucleotide addition and premature termination, respectively (Figure 2(a), lanes 2-3). All three products initiated from the 1 cytidylate, since its change to a guanylate abolished all synthesis ( Figure 2 (a), lanes 4-5). Products directed by several of the mutant endscripts are shown in Figure 2 (a) (lanes 6-15). Approximately 50 mutant endscripts were characterized according to the following scheme. To get an impression of the sequences important for RNA synthesis, a series of three nucleotide transitions starting at the 3 H end up to position 21 were carried out. In general, mutations within the ®rst seven positions relative to the 3 H end of the template have a more severe effect on RNA synthesis than mutations from position 8 to 15 (Figure 2(b) ). Some mutations near the 5 H terminus also decreased RNA synthesis. We have previously analyzed the effect of changes in positions between 17 and 24 and found that sequence within this region is required for ef®cient RNA synthesis in a yet undetermined mechanism (Sivakumaran & Kao, 1999) . We decided to focus on position 2 to 8 because this sequence is likely to be directly involved in the initiation process (Sun et al., 1996; Sun & Kao, 1997a,b) . We consequently tested every possible single nucleotide substitution from position 1 to 9, along with at least one change at position 10 to 21. In contrast to the BMV subgenomic promoter, where changes at key positions drastically reduced RNA synthesis , we found that none of the substitutions prevented RNA synthesis except for a C/G change at the initiation site. However, position 4 to 7 were comparatively sensitive to substitutions. Interestingly, while single nucleotide substitutions of either the 6 or 7 guanylates with cytidylates resulted in low levels of synthesis, replacement of both guanylates with cytidylates in Figure 1 . Analysis of the secondary structure of endscript B2(À)26G. (a) Visualization of RNA conformations by electrophoresis through a non-denaturing 12 % polyacrylamide gel. Between 30 and 60 mM of RNA puri®ed by denaturing gel electrophoresis were heated to 90 C for two minutes, then allowed to cool over a 30 minute period to room temperature. The samples were then electrophoresed on a 17 cm  15 cm  0.8 cm gel at 200 V until the bromophenol blue migrated to the middle of the gel. The RNAs are SL13, an RNA that exists in equilibrium between an intramolecularly folded hairpin and an intermolecularly folded dimer (lane 1), B2(À)26G (lane 2), 3 U/A (lane 3), 6,7 G/C, 9G/A. (b) The secondary structure of B2(À)26G as predicted by the mfold program. (c) One-dimensional NMR spectrum of the imino protons from B2(À)26G taken at 20 C by a Bruker 600 mHz spectrometer. B2(À)26G, at 0.7 mM, was puri®ed and prepared in 10 % deuterium/90 % water as described in Experimental Procedures. (d) The two-dimensional NOESY spectrum of a 2.0 mM B2(À)26G sample taken at 20 C taken from a Bruker 500 mHz spectrometer. The identity of the nucleotide responsible for the signals in the diagonal plane is indicated on the left-hand side of the Figure. Where noted with a gray spot, the origin of the signal was not determined with certainty and could have arisen from nucleotides in the loop of unbase-paired regions. The vertical and horizontal lines were added to the output containing the NMR signals to identify the connectivity of two nucleotides. The two crosspeaks associated with U4 and indicated by squares were observed in spectra with different mixing time than the one shown. endscript 6,7 G/C restored synthesis to 75 % of that of B2(À)26G (Figure 2(b) ). The mutations in B2(À)26G include some that are expected to affect the two stems (A and B), the 3 nt loop, and the mismatched region within the predominant structure of B2(À)26G (Figure 1(b) ). Mutations that should affect at least a portion of the B2(À)26G structure while retaining the ability to direct synthesis are summarized in Figure 3 (a). A change to an adenylate (3U/A) resulted in RNA synthesis of 80 % despite a predicted destabilizing effect on stem A (Figure 2 (a), lanes 6-7, 2(b), and 3(a)). A change of the 9 guanylate to an adenylate (9 G/A), expected to weaken stem B, resulted in 73 % synthesis while changes at positions 10 and 14 expected to weaken stem B resulted in increased synthesis (Figure 3(a) ). The results indicate that the RNA secondary structure of B2(À)26G can be changed without debilitating plus-strand RNA synthesis by the BMV replicase. Several mutations that are expected to preserve the secondary structure in B2(À)26G were found to have a detrimental effect on RNA synthesis. A change of 2 adenylate to a guanylate, which should retain a wobbled G Á U base-pair instead of a Watson-Crick AÁU base-pair, resulted in 8 % synthesis (Figures 2(b) and 3(b) ). A transition of the 3 uridylate to a cytidylate resulted in 18 % synthesis, while a replacement of three base-pairs in stem A with three GÁ C base-pairs abolished synthesis. Changes in loop 1 also greatly reduced RNA synthesis (Figure 3 (b)). Lastly, since nucleotide substitutions that disrupted stem A had detrimental effects on RNA synthesis, we made a compensatory change to the 6 G/C endscript by replacing the 18 cytidylate with a guanylate. This double mutant, SS1, should maintain a predicted secondary structure similar to B2(À)26G. However, RNA synthesis from SS1 was at 21 %. A similar change of 7 G/C along with the complementary change at 17 C/G in endscript SS2 resulted in only 26 % of the activity of B2(À)26 (Figure 4(a) ). Consistent with the ®ndings that mutations may alter B2(À)26G structure without signi®cant loss of activity (Figure 3 (a)), these results indicate that the preformed, predominant secondary structure of B2(À)26G is not essential for directing plus-strand RNA synthesis. The structures predicted by the mfold program for select mutant endscripts were examined to better understand the RNA determinants in B2(À)26G required for initiating RNA synthesis. The predicted secondary structures of the endscripts directing the initiation of BMV RNA1 and RNA3 Figure 2 . Effect of nucleotide changes on plus-strand RNA synthesis. (a) An autoradiogram of RNAs produced by select endscripts. The reaction products were separated by denaturing 12 % PAGE and visualized by autoradiography. é represents the products of a control reaction with no added template. WT denotes the wild-type sequence of B2(À)26G. Changes to the B2(À)26G template are indicated above the autoradiogram of the RdRp products. CM* denotes an endscript with a 3-5 substitutions (UUU changed to CCC) plus a substitution of position 19-21 (AAG changed to GGG). Endscripts that have an initiation-competent cytidylate at the 1 position are indicated by a , while endscripts with an initiation incompetent 1 guanylate are indicated with a À. The predominant product is 26 nt; lesser amounts of a 27 nt product and a 25 nt product are also observed. (b) Summary of the mutational analysis. The wild-type sequence is written near the top of the Figure. The changes at each of the positions are indicated on the left of the Figure. The effect of the nucleotide substitutions on RNA synthesis is denoted as a percentage relative to the amount of synthesis directed by the B2(À)26G wild-type. All results presented are from at least three independent trials with a standard deviation 411 %. Boxes left blank indicate that the effect of that nucleotide substitution was not tested. are shown along with several derivatives of B2(À)26G (Figure 4 (a)). All endscripts capable of directing RNA synthesis tend to have either an exposed 3 H initiation nucleotide, or have it present in a relatively weak portion of a stem that is rich in A Á U and GÁ U base-pairs. Also, all except B3(À)26G possess a loop region rich in pyrimidines. If such a loop motif is necessary for RNA synthesis, it is not suf®cient since SS1, SS2, and several other mutant endscripts that retain such a motif are debilitated for RNA synthesis. Finally, we note that while the RNAs are predicted to vary in the lengths of the stems from six (assuming a uridylate-uridylate base-pair in several of the endscripts) to nine base-pairs, they are all capable of directing RNA synthesis (Figure 4 (a)). To examine whether the predicted stability of the secondary structure is correlated to the ability to direct RNA synthesis, a plot of the calculated ÁG from the mfold program for all mutants endscripts were plotted against their ability to direct RNA synthesis (Figure 4(b) ). Although the calculated ÁG based on the mfold program is only partially accurate due to the lack of thermodynamic information of the loop region, we should be able to determine whether there is an overall correlation between stability of endscript secondary structures and the ability to direct RNA synthesis. Two extremely unstable endscripts did result in inef®cient RNA synthesis (Figure 4(b) ). Nonetheless, no consistent correlation between RNA stability and the ability to direct RNA synthesis was observed. For example, endscript 10 C/U (ÁG 3.4 kcal/mol) is predicted to have lower stability than B2(À)26G (ÁG 3.9 kcal/mol), but was nonetheless more ef®cient in directing RNA synthesis. In addition, 8 U/G (ÁG 8.8 kcal/mol), was observed to direct RNA synthesis nearly twice as ef®ciently as B2(À)26G. Hence the stability of the RNA second-ary structure was not a consistent indicator of the ability to direct synthesis. Our mutational results are inconsistent with the possibility that the interaction between the BMV replicase and B2(À)26G is primarily determined by the predominant solution structure of the RNA. Therefore, we wanted to perform in-depth analyses of select endscripts using NMR and RNA melting experiments. Three mutant endscripts were selected from those described in Figure 2 (1) 3U/A contains a change of the 3 uridylate to adenylate; (2) 6,7 G/C has both the 6 and 7 guanylates changed to cytidylates; and (3) RNA 9 G/A has the 9 guanylate changed to an adenylate. Although each mutation should affect the stability of the different regions within B2(À)26G, the ability to direct RNA synthesis was at least 70 % of that of B2(À)26G (Figure 3 (a)). Native gel electrophoresis was performed to examine the structures of the three mutant endscripts. In a 12 % native gel, each of the three mutant endscripts exists predominantly in a single electrophoretic conformation (Figure 1(a) ). However, the mobility of the three mutant RNAs differ from each other and from that of B2(À)26G. The mobility of 3 U/A, 9 G/A, and 6,7 G/C in comparison to B2(À)26G were, respectively: slightly reduced, drastically reduced, and increased ( Figure 1(a) , lanes 2-5). In a denaturing gel, all the three RNAs co-migrated with B2(À)26G. These results suggest that the differences in mobility of the RNAs are likely due to changes in the RNA conformation. Next, we examined the melting pro®les of the RNAs. The shape of the melting curve for B2(À)26G, 3 U/A and 9 G/A suggests that they melted with multiple step transition, while 6,7 G/C undergoes a single step transition between two states ( Figure 5(a) ). With a crude estimate of the relative increase in absorbance and the magnitude of the slope for each transition, it seems that B2(À)26G, 3U/A and 9G/A underwent a relatively sharp major transition and then a broad minor transition at higher temperatures ( Figure 5(a) ). The multiple transitions might be due to differential melting of secondary structure elements or the unstacking of the bases in the single-stranded portions of the RNAs. The absence of multiple transitions in 6,7G/C clearly shows that it has distinct thermodynamic properties from the other endscripts, implying a different conformation. In addition, even among B2(À)26G, 3 U/A and 9 G/A, the overall shapes of the melting curves look different, which indicates that these RNAs undergo different unfolding processes. These results are consistent with the native gel analysis and suggest that the three mutant endscripts are altered in their secondary structures in comparison to B2(À)26G and to each other. To characterize the changes in the structures of the mutant endscripts more precisely, 1D imino spectra were taken at different temperatures ( Figure 5(b) ). The quality of the ca 12 peaks characteristic of B2(À)26G were signi®cantly affected by temperature. The peaks were relatively sharp at 20 C (data not shown), but became increasingly broad at 30 C and 40 C, indicating that exchange with the environment of some of the imino protons increased at higher temperatures ( Figure 5(b) ). For ease of description, several of the imino peaks have been named in the spectra of B2(À)26G ( Figure 5(b) ). Of interest is the U8-U16 interaction at approximately 11 and 9.7 ppm. This interaction appeared to be more sensitive to temperature changes compared to interactions associated with G ÁC base-pairs at approximately 13-13.4 ppm. Endscript 3U/A has only ®ve of the peaks associ-ated with B2(À)26G ( Figure 5(b) ). At 20 C, the U8-U16 interaction was retained in 3 U/A, as were the peaks associated with G Á C base-pairs adjacent to the U8-U16 interaction. However, missing were the base-pairs at the bottom of stem A, as exempli-®ed by the absence of the U5 peak at 14.4 ppm. At 40 C, all the peaks were signi®cantly broadened in comparison to those from B2(À)26G, indicating that the mutation in 3U/A RNA destabilizes the structure of the RNA. Endscript 6,7 G/C gave a number of sharp peaks that do not correspond to the ones observed in B2(À)26G ( Figure 5(b) ), indicating that 6,7 G/C exists in a different structural conformation than B2(À)26G. The structure predicted for 6,7 G/C by the mfold program was different from that for B2(À)26G (Figure 4(b) ). In addition, since the peaks from 6,7 G/C remain well de®ned at 40 C, its conformation is more (Jaeger et al., 1989) . These structures were selected because they contain stems and loops that are useful to compare with those from B2(À)26G. The amount of synthesis directed by each endscript is shown as a percentage relative to that of B2(À)26G. The ÁG in kcal/mol of the predicted structure was calculated by the mfold program. Endscripts SS1 and SS2, (within the box), denote B2(À)26G RNA derivatives with compensatory nucleotide changes on both sides of the main stem. These were the only two mutant derivatives of B2(À)26G shown here that were signi®cantly debilitated for RNA synthesis. The Other RNA derivatives are described in the legend to Figure 2 (b) and in the text. (b) A plot of relative RNA synthesis versus the calculated ÁG value of the predicted secondary structures for all of mutant endscripts tested in this study. The value for B2(À)26G is denoted by a white diamond and the two endscripts with the least and most stable predicted secondary structures able to direct ef®cient RNA synthesis are indicated by arrows. stable than that of B2(À)26G. This observation is consistent with its UV melt properties ( Figure 5(a) ). Endscript 9 G/A RNA, like 3 U/A RNA, retained a subset of the imino peaks observed in B2(À)26G, mostly corresponding to those in stem A (Figures 1(d) and 5(b) ). However, these peaks are different from those observed in 3 U/A RNA. Most notably, the U8-U16 signal was drastically broadened, indicating the absence of the U8-U16 interaction adjacent to the mutation at position 9. Compared with the spectrum taken at 20 C (data not shown), most peaks were shifted and broadened signi®cantly at 30 C ( Figure 5(b) ). At 35 C, all the peaks were severely broadened, indicating a rapid exchange of the imino protons with the environment due to destabilization of its conformation. Taken together, the biophysical characterization strongly suggest that the predominant structure in B2(À)26G RNA was signi®cantly altered by the introduced mutations. Therefore, the ability of endscripts to direct RNA synthesis by the BMV replicase under conditions that affect RNA conformation is of interest. We examined the effect of temperature on RNA synthesis using B2(À)26G and the three mutant endscripts. Since temperature should affect both the RNA conformation and the activity of the BMV replicase, all syntheses are normalized to that of B2(À)26G at each assay temperature. While increasing the temperature from 30 C to 43 C did result in reduced synthesis (Figure 6) , none of the mutant endscripts were differentially affected in comparison to B2(À)26G. Altogether, these biochemical and biophysical results indicate that multiple RNA secondary structures are acceptable for ef®ciency of RNA synthesis by the BMV replicase. These results also suggest that the template sequence may contribute more signi®cantly to the level of RNA synthesis than its speci®c RNA structure. Mutations in B2(À)26G could affect RNA synthesis either through modulation of the polymerization process, or by preventing template-replicase interaction. To examine the latter possibility, competition assays were performed with several endscripts. This assay determines the ability of competitor RNAs to reduce synthesis directed by an ef®ciently recognized subgenomic RNA, À20/15 Sivakumaran & Kao, 1999) . Competitor RNAs were added at 10 and 20 molar excess of À20/15 and the amount of synthesis directed by À20/15 was quanti®ed. As a Figure 5 . Characterization of the structures of endscript B2(À)26G and three mutant derivatives. (a) Change in absorbencies (at A 260 ) as a function of temperature. The absorbencies were recorded as described in Experimental Procedures. The identities of the RNAs are labeled next to the plot for each of the RNAs. (b) One dimensional NMR spectra of B2(À)26G and three mutant derivatives. The RNAs used ranged from 0.2 mM to 0.7 mM and the spectra were taken with a Bruker 600 mHz spectrometer. Similar regions of the spectrum from the samples are shown. The identities of several of the peaks were determined using 2D-NOESY data shown in Figure 1 (d) and indicated above the peaks. All of the spectra in the left panel were taken at 30 C. The spectra in the right-hand panel were taken at 40 C for B2(À)26G, 3U/A, and 6,7G/C, and at 35 C for 9G/A. Figure 6 . Effect of temperature on the ability of B2(À)26G and mutant endscripts to direct RNA synthesis. RdRp assays were carried out at 30 C, 35 C, 40 C and 43 C using B2(À)26G, 3 U/A, 6,7 G/C, and 9 G/A as templates. The amount of RNA synthesis directed by the various endscripts are presented as a percentage compared to B2(À)26G. All results presented are from at least three independent trials and the range for one standard deviation shown by the vertical lines. negative control, we tested RNA B2(À)Á3 H , which has the three 3 H nucleotides deleted. In comparison to a reaction lacking a competitor, B2(À)Á3 H reduced synthesis from À20/15 to 86 % at ten molar excess and to 76 % at 20 molar excess (Table 1) . This mild reduction in À20/15 synthesis is consistent with our previous observations that the initiation site is required for recognition by the BMV replicase (Sivakumaran & Kao, 1999 ). In contrast, three functional RNAs, B1(À)26G, B2(À)26G, and B3(À)26G encoding the 3 H ends of BMV RNA1, RNA2, and RNA3, reduced synthesis from À20/15 more severely, to 52, 63, and 63 % at ten molar excess, respectively (Table 1) . CMV2(À)26G, the equivalent 3 H end of minus-strand RNA2 of CMV, reduced synthesis to 49 %, a more effective level of inhibition than seen with BMV RNAs (Table 1) . While CMV2(À)26G has the same 3 H three nucleotides as BMV RNAs and can interact with the BMV replicase, it was unable to direct ef®cient synthesis by the heterologous BMV replicase (Sivakumaran & Kao, 1999) . Endscripts 3 U/A, 9G/A, 6,7 G/C, and B3(À) 4U/C were at least as capable competitors as B2(À)26G (Table 1) . These results suggest that the different levels of plus-strand RNA synthesis observed from the different endscripts is not due to a lack of recognition by the BMV replicase. Although it is possible that some RNA synthesis may be negatively affected by a replicase-endscript interaction that is too tight. Nonetheless, these competition results suggest that polymerization of nucleotides following the initial binding of the replicase to the initiation sequence contributes to the ef®ciency of RNA synthesis. We had previously reported that the BMV replicase could utilize DNA templates (Sivakumaran & Kao, 1999; Siegel et al., 1999) . Furthermore, guanylates present within the ®rst four positions of the initiation cytidylate were found to be detrimental to RNA synthesis from DNA endscripts (Sivakumaran & Kao, 1999) . In our mutational analysis we observed that guanylates introduced near the initiation cytidylate of RNA B2(À)26G had a similar inhibitory effect (Figure 2(b) ). Replacement of the native nucleotide with a guanylate from the 2 to 5 positions allowed synthesis between 8 to 44 % of B2(À)26G. The detrimental effect of guanylates decreased as distance from the initiation cytidylate increased (Figure 7) . From position 6 to 14, guanylates did not signi®cantly affect RNA synthesis and sometime increased synthesis, as evidenced by guanylates at position 8 and 10 (Figure 7) . However, guanylates at position 15 to 18 reduced RNA synthesis (Figure 7) . The changes within position 15 to 18 likely affected a sequence previously observed to be important for RNA synthesis (Sivakumaran & Kao, 1999) . Altogether, these results suggest that the identity of nucleotides at speci®c positions may affect the ef®ciency of RNA synthesis by the BMV replicase in vitro. To determine whether guanylates near the initiation site have a possible effect in vivo, we examined the distribution of guanylates in the sequences of BMV and other representatives of the alphavirus family. Sequences immediately following the initiation of genomic and subgenomic plusstrand RNAs generally lacked a guanylate in the ®rst four positions (Table 2) . At the 5 position, some animal-infecting, but no plant-infecting RNA viruses, had a guanylate in their RNAs. Guanylates were then found in increasing frequency starting at the 6 position. The absence of template guanylates near the initiation nucleotide may be due to a requirement for replicase-promoter recognition, or because the interaction of the nascent RNA with the template is affected by tighter base-pairing. In the latter case cytidylates near the initiation site should also be detrimental to synthesis. While we did observe that substitutions with cytidylates near the initiation cytidylate did have detrimental effects, the effects were not to the same degree as the guanylate substitutions. However, in examining the 3 H ends of genomic plus-strand templates of approximately 60 viral templates, we observed a similar striking bias against guanylates and cytidylates from position 2 to 5, followed by a more normal distribution of these two nucleotides (Figure 8) . This trend appears to be unique for templates directing the synthesis of plus-strand RNAs, since minus-strand synthesis does not follow the same trend ( Figure 8) . Together, the phylogenetic and mutational results are consistent with the testable hypothesis that the strength of the interaction between the template and the nascent RNA near the initiation site contributes to the ef®ciency of plus-strand RNA synthesis. The observation suggesting that the primary sequence of BMV RNAs can affect RNA synthesis prompted us to examine the BMV RNA sequences more closely. The templates for synthesis of BMV RNA2 and RNA3 are identical in the ®rst ®ve pos-itions and RNA1 and RNA2 are identical at the ®rst 26 positions except at 4, where RNA1 has a cytidylate and RNA2 a uridylate (Figure 9(a) ). In vitro, B1(À)26G directs only 42 % the amount of product as B2(À)26G (Figure 9(b) , lanes 1-2 and 3-4). In vivo, RNA1 is known to accumulate at 68 % of RNA2 (Marsh et al., 1991) . To determine whether the 4 position is important in regulating the ef®ciency of RNA synthesis, we made B3(À)26G with a 4U/C change and quanti®ed the amount of synthesis relative to B2(À)26G. Synthesis directed by B3(À)26G was 114 % of that from B2(À)26G (Figure 9(b) , lanes 1-2 and 5-6). B3(À)26G with a 4 U/C change decreased synthesis to 52 % of that from B3(À)26G (Figure 9(b) , lanes 5-6 and 7-8). These results provide another example that the identity of a template nucleotide near the initiation site can affect the ef®ciency of RNA synthesis. An understanding of the rules guiding RNA synthesis is lacking in all RNA viruses. Determining these rules will provide powerful guides for developing strategies to inhibit viral infections and to compare the mechanism of RNA synthesis by other polymerases. We have previously demonstrated that initiation of BMV genomic plus-strand synthesis can be accurately directed by minimal -----1 , 2 1 , 2 -1,2,3 3 --3 1,2,4 -CCMV a -----1,2,3 1,2 -1,2 4 3 3 --1,2,4 CMV a -----3 --4 1,2,3 3 4 3 BMV, brome mosic virus; CCMV, cowpea chlorotic mottle virus; CMV, cucumber mosaic virus (fny); AMV, alfalfa mosaic virus; BBMV, broad bean mottle virus; TMV, tobacco mosaic virus; EEE, eastern equine encephalitis virus; VEE, Venezuelan equine encephalitis virus; RRV, Ross River virus. The numbers 1, 2, 3 represent genomic RNA1, 2, 3 respectively; 4 represents the subgenomic RNA encoding the capsid protein, and 5 represents the 30 K subgenomic RNA encoding the movement protein of TMV. a Ahlquist et al. (1981) . b Gunn & Symons (1980) . c Goelet et al. (1982) . d Ou et al. (1982a Ou et al. ( ,b, 1983 ). e Rumenapf et al. (1995) . f Faragher & Dalgarno (1986) . Faragher et al. (1988) . g Shirako et al. (1991) . h Levinson et al. (1990) . minus-strand templates in vitro (Sivakumaran & Kao, 1999) . A more detailed characterization of a functional endscript of 27 nt, B2(À)26G, was undertaken in the present study. In this work we demonstrated that: (1) the preformed RNA secondary structure of B2(À)26G is not speci®cally required for template recognition by the BMV replicase; (2) the RNA promoter for RNA synthesis requires nucleotides À1, 1 and 2 as a determinant for replicase binding; (3) template sequence beyond the 3 position likely affects RNA synthesis by modulating the polymerization process; and (4) phylogenetic analysis of the viral sequences indicate that the in vitro requirements described likely re¯ect in vivo requirements. The results are placed below in context of the mechanism of viral RNA replication. The BMV subgenomic promoter is recognized by the BMV replicase in a sequence-speci®c manner . Results herein demonstrate that the endscript directing genomic plus-strand synthesis also interacts with the viral replicase via a mechanism that does not depend on a speci®c preformed RNA structure. For the subgenomic promoter, the recognition nucleotides include the initiation cytidylate (1) and at least four positions 3 H of the initiation site (positions À17, À14, À13, and À11) in a region determined to be important for BMV subgenomic RNA synthesis in infected cells (French & Ahlquist, 1988) . For the genomic endscript, interaction with the replicase requires the initiation cytidylate and the neighboring À1 and 2 nucleotides. The sequence 3 H to the subgenomic core promoter has been demonstrated to modulate the level of RNA synthesis (Marsh et al., 1988; Adkins et al., 1997) . In the genomic endscript, 5 H sequence from 17 to 24 affect the level of RNA synthesis. Finally, both templates are rich in adenylates and uridylates. With the genomic endscript the A U-rich nature of the template could affect synthesis, but not binding to the replicase (Figure 2(b) and Table 1 ). The sequences directing plus-strand syntheses are part of the intermediate synthesized during viral replication. Given suf®cient time, these intermediates will either associate intermolecularly with the complementary template to form stable double-stranded RNAs, or fold intramolecularly into a highly structured state (unless there are proteins that affect RNA structure). Since there are no sequences in common for genomic plus and subgenomic RNA synthesis except for the initiation cytidylate, it is likely that both promoters are recognized before stable RNA structures can be formed in the nascent RNA. Based on these and other observations (Bujarski et al., 1994; Adkins Figure 8 . A selection of 60 plus-strand viral RNA templates reveals a propensity against guanylate and cytidylate proximal to the genomic plus-strand initiation site. Genomic minus-strand initiation sequences of the same templates do not exhibit such a preference. Sequence data were obtained from NCBI Entrez. Only plus-sense RNA viruses with complete genomic sequences and with plant tropism were included. For database accession numbers, see Experimental Procedures. Figure 9 . Effect of nucleotide changes on the differential accumulation of BMV RNAs. (a) Comparison of the 3 H minus-strand RNA sequences of BMV. The non-templated guanylate at the 3 H end is shown in lower case g. The arrow denotes the initiation cytidylate. The 4 U/C change in B3(À)26G endscript is shown in bold. (b) RNA synthesis directed by B1(À)26G, B2(À)26G, B3(À)26G and B3(À) 4 U/C. RdRp reaction products were separated by 12 % denaturing PAGE and visualized by autoradiography. The amount of RNA synthesis from various templates relative to B2(À)26G are summarized at the bottom of the autoradiogram. Each number presented is an average from three independent trials. Siegel et al., 1997 Siegel et al., , 1998 Sivakumaran & Kao, 1999) , we propose that the initiation of BMV genomic and subgenomic plus-strand RNA synthesis takes place while the replicase is still associated with the nascent RNA ( Figure 10 ). In this model, when minus-strand RNA synthesis reaches the BMV RNA3 intercistronic sequence encoding the subgenomic promoter, the replicase is faced with a roadblock that signals the replicase to either continue copying the template RNA or reverse course and de novo initiate subgenomic RNA synthesis from the nascent RNA. Overcoming this roadblock may depend on the ability of the replicase to successfully pass the intercistronic sequence, which in several bromoviruses contain an unusual homopolymeric sequence (Marsh et al., 1988; Allison et al., 1988 Allison et al., , 1989 Romero et al., 1992) or contain stable stem-loop structures in other viruses (Zavriev et al., 1996; Koev et al., 1999) , or stable RNA-RNA interactions (Kim & Hemenway, 1999; Sit et al., 1998) . The replicases that cannot pass the intercistronic sequence would release the template RNA and initiate subgenomic RNA synthesis from the recently synthesized nascent RNA. The precise initiation position would be determined by the speci®c key nucleotides present in the subgenomic core promoter . Bujarski et al. (1994) have noted that the BMV intercistronic sequence is a region for frequent template switching, and Nagy et al. (1999) have demonstrated that the AU content of a template may promote recombination, a process that depends on template switching. Further support for this hypothesis include the observations by French & Ahlquist (1988) , who by placing a series of BMV subgenomic promoters at different positions found that promoters closer to the start site of minus-strand RNA synthesis directed more subgenomic RNA products. The replicase molecules that could successfully pass through the intercistronic sequence would proceed forward generating full-length minusstrand RNAs. Upon reaching the end of the viral template, the replicase adds a non-templated nucleotide to the nascent minus-strand RNA by a terminal transferase activity . The recognition of the initiation site (nucleotides À1, 1 and 2) signals the replicase to initiate genomic plus-strand synthesis from the nascent minus-strand RNA template. The differential accumulation of subgenomic and genomic RNAs in vivo (Marsh et al., 1991) could be due to a larger number of replicase molecules stalling at the intercistronic promoter region and reversing course to initiate subgenomic RNA synthesis. The hypothesis that plus-strand RNAs are synthesized from the nascent minus-strand RNA without releasing the polymerase has been proposed for several viral systems. In Coronaviruses, minusstrand RNAs approximately the length of subgenomic RNAs were produced during infection (Sethna et al., 1989) and these replication intermediates were found to be transcriptionally active (Stawicki & Stawicki, 1990) , suggesting that incompletely synthesized minus-strand RNA may direct plus-strand RNA synthesis. In the polycistronic red clover necrotic virus, the interaction in trans between two RNA sequences may prevent the progression of the viral replicase and induce subgenomic RNA synthesis (Sit et al., 1998) . For the polycistronic potato virus X RNA, a series of repeated sequences that resemble transcription pause/termination sites of DNA templates may act to slow the progression of the replicase during minus-strand synthesis and thereby induce the initiation of plus-strand RNA synthesis (Kim & Hemenway, 1999) . The sequence-speci®c requirements for plusstrand genomic and subgenomic syntheses are Figure 10 . A model for the synthesis of genomic and subgenomic plus-strand RNAs by the BMV replicase. Schematic representation of BMV RNA3 (not intended to be at correct proportions) with the coding sequences denoted by long rectangles. The untranslated 5 H , 3 H and the intercistronic region are shown as lines. The replicase is shown as an oval with the direction of synthesis indicated by an arrow. The replicase binds to the 3 H tRNA-like end of genomic RNA and initiates synthesis of the minus-strand RNA. The nascent strand is shown attached to the replicase. Upon reaching the intercistronic region the replicase could either continue minusstrand synthesis or reverse course to initiate subgenomic plus-strand RNA synthesis in a process that require replicase to use nascent minus-strand as template. The replicase could also pass through the homopolymeric region and complete minus-strand synthesis. Upon reaching the end of the template, the replicase adds a non-templated nucleotide by terminal transferase activity. The 3 H non-templated nucleotide, along with the initiation nucleotide signals the replicase to switch to the nascent-strand to initiate genomic plus-strand synthesis. apparently in contrast to the requirements for genomic minus-strand synthesis. Structures at the 3 H end of the viral genomic RNA that are required to direct genomic minus-strand synthesis have been demonstrated in several systems, including BMV (Dreher & Hall, 1988; Pogue & Hall, 1992; Philipenko et al., 1996; Song & Simon, 1995; Draper, 1995; Hong et al., 1998; Chapman & Kao, 1999) . There may be major differences in the mode of recognition of promoters for minus versus plusstrand RNA synthesis. For plus-strand RNA viruses the genomic RNA is present in the infected cell prior to its translation and assembly of the replicase. Since all molecules prefer to exist in a lower energy state, the viral genomic RNA would possess highly ordered secondary and tertiary structures, which may be formed with the assistance of sequence-non-speci®c RNA-folding chaperons or sequence-speci®c RNA-binding cofactors (Weeks, 1997) , such as the BMV 1a protein (Sullivan & Ahlquist, 1999) . Hence, the replicase needs to recognize the highly structured genomic RNA to initiate minus-strand synthesis. The differences in the incidence of guanylates and cytidylates ( Figure 8 ) provide another indication of the differences between minus and plus-strand RNA synthesis. The initiation of minus-strand RNA synthesis may also be affected by upstream sequence (Carpenter et al., 1995; Sullivan & Ahlquist, 1999; Chapman et al., 1998) . An interesting conclusion from our analysis of the BMV genomic minus-strand RNA endscript is that template sequence may regulate the amount of RNA synthesis in a manner independent of replicase-promoter interaction. This is different from the general notion of promoter-strength, which implies a modulation of the frequency of initiation events, and necessitates an emphasis on the frequency of successful transitions into processive elongation. Transcription studies using DNA templates indicate that the template sequence could affect the proportion of RNAs that are successfully transcribed to completion (Keene & Luse, 1999; Nudler et al., 1997; Von Hippel & Yager, 1991) . A similar strategy may regulate RNA synthesis from RNA templates. An assumption of this hypothesis is that the viral sequence contains nucleotides that are selected at speci®ed positions to enable proper transition from initiation to elongation phase of RNA synthesis and the optimal level of RNA accumulation. The difference in BMV RNA1 and RNA2 at the 4 position provides an example where a single-nucleotide change in the viral RNA can account for a difference in RNA accumulation (Figure 9 ). The BMV replicase is known to increase its af®nity for interaction with the template RNA as synthesis progresses (Sun & Kao, 1997a,b) . The increased af®nity may affect the stability of the template-nascent RNA duplex interaction, i.e. too stable an interaction during initiation may prevent the replicase from making the transition from the initiation site, resulting in decreased RNA synthesis. Consistent with this testable hypothesis, we observed that guanylates and cytidylates negatively affected RNA synthesis when they were within four positions of the initiation nucleotide. Inconsistent with this hypothesis, however, is the fact that substitutions with guanylate and cytidylates do not have identical effects (Figure 2(b) ). Stability of the template-nascent strand duplex is also affected by the contributing effects of the neighboring nucleotides (Turner et al., 1988) , and a cytidylate substitution at a given position may not have the same thermodynamic effect as a guanylate. Alternatively, thermodynamic parameters may not be the sole factor affecting viral RNA synthesis. Another possible effect could be that the template sequence might also have evolved to in¯uence translation. The transition from initiation to elongation phase during BMV RNA synthesis takes place at positions 8 to 10 (Sun & Kao, 1997a,b; . The observation that guanylates and cytidylates are naturally enriched between position 6 to 10 suggests that these positions are acting in a manner different than the positions required for initiation. Perhaps a tighter template-nascent RNA interaction is required for ef®cient transition from initiation to elongation (Figure 2(b) ). Consistent with this, the introduction of a guanylate at position 8 or 10 in B2(À)26G increased RNA synthesis above wild-type levels (Figure 2(b) ). Our results indicating that viral replicase-promoter interaction being independent of the initial RNA secondary structure may be interpretable by two general models. First, the recognition of the promoter may be directed by the sequence of the RNA. Sequence-speci®c recognition has been previously demonstrated for the bacteriophage T4 translational repressor RegA protein (Webster & Spicer, 1990; Brown et al., 1997) . Second, it is possible that the recognition of the genomic minusstrand RNA by the viral replicase could take by inducing tertiary interaction between protein and RNA which is tolerant of some differences in the RNA secondary structure. An induced-®t mechanism has been suggested for a number of RNAbinding proteins, including the splicing factor, U1A, the ribosomal S15 protein, and the binding of the hepatitis B encapsidation signal by the P protein (Gubser & Varani, 1996; Batey & Williamson, 1996; Beck & Nassal, 1997) . There is also evidence for an induced ®t mechanism for the recognition of the BMV subgenomic promoter by the BMV replicase. However, in our studies, the contacts in the RNA will induce a conformational change in the replicase, resulting in different nucleotide moieties being recognized Stawicki & Kao, 1999) . A subsequent change in the RNA sec-ondary or tertiary structure after replicase binding to the RNA has not been ruled out. This induced ®t mechanism is compatible with the ability of the replicase to recognize the subgenomic and genomic promoters without the release of the replicase during minus-strand RNA synthesis. Wild-type and mutant endscripts were prepared for NMR analysis by in vitro transcription using phage T7 RNA polymerase and synthetic DNA templates (Milligan et al., 1987; Wyatt et al., 1991) . The transcribed RNAs were puri®ed by denaturing polyacrylamide gel electrophoresis and eluted from the gel slices using a electroelution trap (Scheicher & Schuell). The eluted RNA was concentrated by ethanol precipitation and dialyzed against the NMR buffer (10 mM sodium phosphate, 0.1 mM EDTA, 100 mM NaCl, pH 6.5) for 48 hours. Finally, the samples were dried down and dissolved in the same volume of 90 % H 2 O/10 % 2 H 2 O. The RNA samples are at 0.4 to 1 mM for 1D spectra and 2.0 mM for 2D H 2 O NOESY experiment. Exchangeable proton 1D spectra were recorded on a Bruker AMX-600 spectrometer equipped with a Bruker triple resonance triple-axis gradient probe. Solvent suppression was achieved using the jump-return method with the excitation maximum set around the middle of the usual chemical shift range of the imino protons (12.5 ppm) (Plateau & Gueron, 1982) . A spectral width of 12,000 Hz was used and 4096 complex data points were collected with a relaxation delay of two seconds. Depending on the sample concentration, 32 to 1024 scans were taken. In analysis of the mutant endscripts, all of the gains were set at the same level. The 2D H 2 O NOESY spectra were collected independently using a Bruker AMX-600 spectrometer equipped with a Bruker triple resonance triple-axis gradient probe and a Bruker DRX-500 spectrometer equipped with a Nalorac triple resonance z-axis gradient probe. The results from the two spectrometers were consistent. H 2 O NOESY spectra were collected at mixing times 200 ms and 300 ms at 20 C. A total of 400 to 440 t 1 blocks were taken, for each block, 32 scans were taken with a relaxation delay of two seconds. Solvent suppression was accomplished by replacing the last pulse in NOESY pulse sequence with a jump-return pulse sequence and applying z-gradient pulse during the mixing time. Proton chemical shifts were indirectly referenced to DSS using the water resonance. All NMR spectra were processed using FELIX 95.0 (Biosym Technologies, Inc.). RNA conformation native gel analysis was performed in 12 % polyacrylamide gels (40:2 acrylamide to bisacrylamide) containing 0.5 TBE (Maniatis et al., 1982) . After electrophoresis for 20 minutes, 6 ml of a 6 mM sample was loaded in a dye containing 0.5 TBE, 5 % (v/v) glycerol and bromophenol blue. Electrophoresis was performed at a constant 200 volts until bromophenol blue migrated for 6-8 cm. The gel was then stained with Toluidine blue to visualize the RNAs. The protocol used was as described by Puglisi & Tinoco (1989) using a Gilford spectrophotometer model 250 and a model 2527 thermoprogrammer. Brie¯y, the sample concentration was determined by calculations using Beer-Lambert Law with a calculated extinction coef®cient for B2(À)26G of 286. After adjusting the absorbance of the samples to be between 0.8 and 1.2 at 20 C, data points were taken with a change in temperature of 0.5 deg. C/minute. The results were analyzed and plotted using a Kaleidagraph software package. t m for each was calculated using a derivative of the raw data. Synthesis of RNA templates for RdRp assay cDNA copies of (À)-strand 3 H ends encompassing the ®rst 26 nt of BMV RNA 1, 2 and 3 were generated by PCR ampli®cation. Pairs of oligonucleotide primers were used, one of which contained a T7 promoter to enable transcription by T7 RNA polymerase. Following transcription, the RNAs were electrophorsed in a denaturing PAGE, and the band of correct molecular mass identi®ed by staining with ethidium bromide and excised with a razor. The gel was then crushed to elute the RNA in a solution of 0.3 M sodium acetate. Concentrations of RNA were determined by spectrometry and staining with Toluidine blue following denaturing PAGE. BMV replicase was prepared from infected barley as described by Sun et al. (1996) . Replicase activity assays were carried out as described by . Brie¯y, each assay consisted of a 40 ml reaction mixture containing 20 mM sodium glutamate (pH 8.2), 12 mM dithiothreitol, 4 mM MgCl 2 , 0.5 % (v/v) Triton X-100, 2 mM MnCl 2 , 200 mM ATP, 500 mM GTP, 200 mM UTP, 242 nM [a-32 P]CTP (400 Ci/mmol, 10 mCi/ml, Amersham), the desired amount of template and 7 ml of RdRp. Following incubation for 90 minutes at 30 C, the reaction products were extracted with phenol/chloroform (1:1, v/v) and precipitated with six volumes of ethanol, 10 mg of glycogen, and 0.4 M ®nal concentration of ammonium acetate. Loading buffer (45 % (v/v) deionized formamide, 1.5 % (v/v) glycerol, 0.04 % (w/v) bromophenol blue, and 0.04 % (w/v) xylene cyanol) was added to the products and denatured by heating at 90 C for three minutes prior to electrophoresis on 12 % acrylamide-7 M urea denaturing gels. All gels were dried and exposed to ®lm at À80 C and the amount of label incorporated into newly synthesized RNAs was determined with a Phosphorimager (Molecular Dynamics). The template sequences proximal to the plus and minus-strand initiation sites were found at the NCBI website (http://www.ncbi.nlm.nih.gov). Viral sequences used were those indicated as complete by the Entrez database. The species and accession numbers are: alfalfa mosaic virus RNA1 (L00163 J02000), alfalfa mosaic virus RNA2 (K02702 J02002), alfalfa mosaic virus RNA3 (K03542), apple chlorotic leaf spot virus (D14996), apple stem grooving virus (D14995 S47260), barley mild mosaic virus RNA1 (AJ242725), barley mild mosaic virus RNA2 (X75933), beet necrotic yellow vein mosaic brome mosaic virus RNA2 (X01678 K02707), brome mosaic virus RNA3 (J02042 J02043), carnation mottle virus (X02986), carrot mottle mimic umbravirus RNA1 (U57305), cowpea chlorotic mottle virus RNA1 (M65139), cowpea chlorotic mottle virus RNA2 (M28817), cowpea chlorotic mottle virus RNA3 (M28818), cowpea mosaic virus RNA B (X00206), cowpea mosaic virus RNA M (X00729), cucumber mosaic virus RNA1 (D12537 D01199), cucumber mosaic virus RNA2 (D12538 D01200), cucumber mosaic virus RNA3 (D10539), cymbidium mosaic virus (U62963), grapevine virus A (X75433), olive latent virus 1 (X85989), olive latent virus 2 RNA1 (X94346), olive latent virus 2 RNA2 (X94347), olive latent virus 2 RNA3 (X76993), olive latent virus 2 RNA4 (X77115), papaya mosaic virus (D13957 D00580), peanut clump virus RNA1 (X78602), peanut clump virus RNA2 (L07269), peanut mottle virus (AF023848), pepper mottle virus (M96425), potato mop-top virus RNA3 (D16193), potato virus X (D00344), potato virus Y (A08776), raspberry bushy dwarf virus RNA1 (S51557), ryegrass mosaic virus (AF035818), soil-borne wheat mosaic virus RNA1 (L07937), soil-borne wheat mosaic virus RNA2 (L07938), soybean mosaic virus (S42280), sweet clover necrotic mosaic virus (S46028), tobacco mosaic virus (D13438) Subgenomic RNA promoters dictate the mode of recognition by bromoviral RNA-dependent RNA polymerases Minimal templates directing accurate initiation of subgenomic RNA synthesis in vitro by the brome mosaic virus RNA-dependent RNA polymerase Mechanistic analysis of RNA synthesis by RNA-dependent RNA polymerase from two promoters reveals similarities to DNA-dependent RNA polymerase Bromovirus RNA replication and transcription Near identity of 3 H RNA secondary structure in bromoviruses and cucumber mosaic virus Infectious in vitro transcripts from cowpea chlorotic mottle virus cDNA clones and exchange of individual components with brome mosaic virus Sequence of cowpea chlorotic mottle virus RNAs 2 and 3 and evidence of a recombination event during bromovirus evolution Interaction of the Bacillus stearothermophilus ribosomal protein S15 with 16S RNA. II. Speci®city determinants of protein-RNA recognition Sequence and structure-speci®c determinations in the interaction between RNA encapsidation signals and reverse transcriptase of avian hepatitis B viruses Identi®cation and properties of the RNA-dependent RNA polymerase of hepatitis C virus Cosuppression comes to the animals Single-stranded RNA recognition by the bacteriophage T4 translational repressor Comparison of the replication of positive-strand RNA viruses of plants and animals Molecular studies of genetic RNA-RNA recombination in brome mosaic virus Involvement of a stem-loop structure in the location of junction sites in viral RNA recombination A minimal RNA promoter for minus-strand RNA synthesis by the brome mosaic virus polymerase complex Sequences 5 H of the conserved tRNA-like promoter modulate initiation of minus-strand synthesis by the brome mosaic virus RNA-dependent RNA polymerase Gene silencing in Neurospora crassa requires a protein homologous to RNA-dependent RNA polymerase Protein-RNA recognition Mutational analysis of the sequence and structural requirements in brome mosaic virus for minus-strand promoter activity Regions of conservation and divergence in the 3 H untranslated sequences of genomic RNA from Ross River virus isolates Genome sequences of a mouse-avirulent and mouse-virulent strain of Ross river virus Characterization and engineering of subgenomic RNA Nucleotide sequence of tobacco mosaic virus Alpha-like viruses in plants Structure of the polyadenylation regulatory element of the human U1A pre-mRNA 3 H -untranslated region and interaction with the U1A protein Sequence homology at the 3 H -termini of the four RNAs of the alfalfa mosaic virus Highly active template-speci®c RNAdependent RNA polymerase from barley leaves infected with brome mosaic virus Novel structures of two virus-like RNA elements from a diseased isolate of the Dutch elm disease fungus Ophiostoma novo-ulmi Improved predictions of secondary structures for RNA Initially transcribed sequences strongly affect the extent of abortive initiation by RNA polymerase II Long-distance RNA-RNA interactions and conserved sequence elements affect potato virus X RNA accumulation Primary and secondary structural elements required for synthesis of barley yellow dwarf virus subgeomic RNA1 Determination of the complete nucleotide sequence of the genomic RNA of O'Nyong-nyong virus and its use in the construction of phylogenetic trees Mutational analysis of the core and modulator sequences of the BMV RNA3 subgenomic promoter Regulation of ():(À)-strand asymmetry in replication of brome mosaic virus RNA Synthesis of brome mosaic virus subgenomic RNA in vitro by internal initiation on (À)-sense genomic RNA Oligoribonucleotide synthesis using T7 RNA polymerase and synthetic DNA templates Mapping sequences active in homologous RNA recombination in brome mosaic virus: prediction of recombination hotspots Puri®cation, characterization, and comparison of poliovirus RNA polymerase from native and recombinant sources The RNA-DNA hybrid maintains the register of transcription by preventing backtracking of RNA polymerase Analysis of RNAdependent RNA polymerase structure and function as guided by known polymerase structures and computer predictions of secondary structure Sequence studies of several alphavirus genomic RNAs in the region containing the start of the subgenomic RNA The 3 H -noncoding regions of alphavirus RNAs contain repeating sequences The 5 H -terminal sequences of the genomic RNAs of several alphaviruses Ciselements, oriR, involved in the initiation of (À)-strand poliovirus RNA: a quasi-globular multidomain RNA structure maintained by tertiary (``kissing'') interactions Exchangeable proton NMR without base-line distortion, using strongpulse sequences Sequence comparison of ®ve polymerases (L proteins) of unsegmented negative-strand RNA viruses: theoretical assignment of functional domains The requirement for a stem-loop structure in brome mosaic virus replication supports a new model for viral positive-strand RNA initiation Absorbence melting curves of RNA Puri®cation and characterization of brome mosaic virus RNA-dependent RNA polymerase The nucleotide sequence and genome organization of the RNA2 and RNA3 segments in broad bean mottle virus Aura virus is a new world representative of Sindbis-like viruses Isolation of an RNAdirected RNA polymerase-speci®c cDNA clone from tomato Coronavirus subgenomic minus-strand RNAs and potential for mRNA replicons Structure of the Ockelbo virus genome and its relationship to other Sindbis viruses Sequence-speci®c recognition of a subgenomic promoter by a viral RNA polymerase Moieties in an RNA promoter speci®cally recognized by a viral RNA-dependent RNA polymerase Use of DNA, RNA and chimeric templates by a viral RNA-dependent RNA polymerase RNAmediated trans-activation of transcription from a viral RNA Initiation of genomic positive strand synthesis from DNA and RNA templates by a viral RNA-dependent RNA polymerase Requirement of a 3 H -terminal stem-loop in in vivo transcription by an RNA dependent RNA polymerase Coronavirus transcription: subgenomic mouse hepatitis virus replicative intermediates function in RNA synthesis Spatial requirements for promoter recognition by a viral RNA-dependent RNA polymerase A brome mosaic virus intergenic RNA3 replication signal functions with viral replication protein 1a to dramatically stabilize RNA in vivo RNA synthesis by the brome mosaic virus RNA-dependent RNA polymerase: transition from initiation to elongation Characterization of RNA products associated with or aborted by a viral RNA-dependent RNA polymerase Initiation of (À)-strand RNA synthesis catalyzed by the BMV RNA-dependent RNA polymerase: synthesis of oligonucleotides RNA structure prediction Identi®cation of poliovirus polypeptide p63 as a soluble RNAdependent RNA polymerase RNA structure and NMR spectroscopy Antisense globin RNA in mouse erythroid tissues: structure, origin, and possible function Transcription elongation and termination are competitive kinetic processes Characterization of bacteriophage T4 RegA protein-nucleic acid interaction Protein-facilitated RNA folding Synthesis and puri®cation of large amounts of RNA oligonucleotides Expression, puri®cation and partial characterization of HCV RNA polymerase Mapping of the red clover necrotic mosaic virus subgenomic RNA This work would not have been possible without the generous support of Professor Ignacio Tinoco Jr, in whose laboratory the NMR and RNA melting experiments were performed. We thank M. Schmitz, R. Gonzalus, for helpful discussions. Funding was provided in part by a National Health Institute grant GM10840 and Department of Energy grant DE-FG03-86ER60406 to I.T. and by the National Science Foundation grant MCB9507344 to C.K.