Double-stranded RNAs of cucumber mosaic virus and its satellite contain an unpaired terminal guanosine: implications for replication

High efficiency T7 polymerase synthesis of infectious RNA from cloned brome mosaic virus cdna and effects of 5' extensions on transcript infectivity

Addition of a single extra 5' G residue reduces the infectivity of transcripts from cloned BMV cDNA no more than threefold, while addition of 7- or 16-base 5' extensions to RNA3 transcripts dramatically suppresses their biological activity. Moreover, despite resultant alteration of the promoter consensus sequence, fusion of BMV cDNA directly to the initiation site of the canonical phi10 promoter of bacteriophage T7 allows efficient in vitro synthesis of infectious BMV transcripts by T7 RNA polymerase, providing substantial improvement in the ease and reproducibility with which BMV cDNA can be expressed. BMV transcripts without a 5' cap were found to be infectious to barley protoplasts, but at a much lower efficiency than capped transcripts.

Nontemplated terminal nucleotidyltransferase activity of double-stranded RNA bacteriophage phi6 RNA-dependent RNA polymerase

The replication and transcription of double-stranded RNA (dsRNA) viruses occur within a polymerase complex particle in which the viral genome is enclosed throughout the entire life cycle of the virus. A single protein subunit in the polymerase complex is responsible for the template-dependent RNA polymerization activity. The isolated polymerase subunit of the dsRNA bacteriophage phi6 was previously shown to replicate and transcribe given RNA molecules. In this study, we show that this enzyme also catalyzes nontemplated nucleotide additions to single-stranded and double-stranded nucleic acid molecules. This terminal nucleotidyltransferase activity not only is a property of the isolated enzyme but also is detected to take place within the viral nucleocapsid. This is the first time terminal nucleotidyltransferase activity has been reported for a dsRNA virus as well as for a viral particle. The results obtained together with previous high-resolution structural data on the phi6 RNA-dependent RNA polymerase suggest a mechanism for terminal nucleotidyl addition. We propose that the activity is involved in the termination of the template-dependent RNA polymerization reaction on the linear phi6 genome.

Parallels among positive-strand RNA viruses, reverse-transcribing viruses and double-stranded RNA viruses

Viruses are exceptionally diverse and are grouped by genome replication and encapsidation strategies into seven distinct classes: two classes of DNA viruses (encapsidating single-stranded (ss)DNA or double-stranded (ds)DNA), three classes of RNA viruses (encapsidating mRNA-sense ssRNA, antisense ssRNA or dsRNA) and two classes of reverse-transcribing viruses (encapsidating RNA or DNA).

Despite substantial life-cycle differences, positive-strand RNA ((+)RNA) viruses, dsRNA viruses and reverse-transcribing viruses share multiple similarities in genome replication. All replicate their genomes through RNA intermediates that also serve as mRNAs. Moreover, the intracellular RNA-replication complexes of (+)RNA viruses share similarities in structure, assembly and function with the polymerase-containing virion cores of dsRNA and reverse transcribing viruses.

Brome mosaic virus (BMV) RNA-replication factors 1a and 2a^pol and cis-acting template-recruitment signals parallel retrovirus Gag, Pol and RNA-packaging signals in virion assembly: 1a localizes to specific membranes, self-interacts and induces ∼60-nm membrane invaginations to which it recruits 2a^pol and viral RNAs for replication. Therefore, like retroviruses and dsRNA viruses, BMV sequesters its genomic RNA and polymerase in a virus-induced compartment for replication.

BMV and some other alphavirus-like (+)RNA viruses also parallel retroviruses in using tRNA-related sequences to initiate genome replication, and share with dsRNA reoviruses aspects of the function and interaction of their RNA polymerase and RNA-capping enzymes. Emerging results indicate that the genome-replication machineries of these viruses might share other mechanistic features.

Whereas (+)RNA alphavirus-like viruses, dsRNA reoviruses and retroviruses are linked by the above similarities, (+)RNA picornaviruses, dsRNA birnaviruses and reverse-transcribing hepadnaviruses share some distinct features, including protein-primed nucleic-acid synthesis. Such parallels suggest that at least some (+)RNA viruses, dsRNA viruses and reverse-transcribing viruses might have evolved from common ancestors. The transitions required for such evolution can be readily envisioned and some have precedents.

These underlying parallels in genome replication by four of the seven main virus classes might provide a basis for more generalizable or broader-spectrum approaches for virus control.

Despite major differences in the life cycles of the seven different classes of known viruses, the genome-replication processes of certain positive-strand RNA viruses, double-stranded RNA viruses and reverse-transcribing viruses show striking parallels. Paul Ahlquist highlights these similarities and discusses their intriguing evolutionary implications.

Viruses are divided into seven classes on the basis of differing strategies for storing and replicating their genomes through RNA and/or DNA intermediates. Despite major differences among these classes, recent results reveal that the non-virion, intracellular RNA-replication complexes of some positive-strand RNA viruses share parallels with the structure, assembly and function of the replicative cores of extracellular virions of reverse-transcribing viruses and double-stranded RNA viruses. Therefore, at least four of seven principal virus classes share several underlying features in genome replication and might have emerged from common ancestors. This has implications for virus function, evolution and control.

Replication of alfamo- and ilarviruses: role of the coat protein

In the family Bromoviridae, a mixture of the three genomic RNAs of bromo-, cucumo-, and oleaviruses is infectious as such, whereas the RNAs of alfamo- and ilarviruses require binding of a few molecules of coat protein (CP) to the 3' end to initiate infection. Most studies on the early function of CP have been done on the alfamovirus Alfalfa mosaic virus (AMV). The 3' 112 nucleotides of AMV RNAs can adopt two different conformations. One conformer consists of a tRNA-like structure that, together with an upstream hairpin, is required for minus-strand promoter activity. The other conformer consists of four hairpins interspersed by AUGC-sequences and represents a strong binding site for CP. Binding of CP to this conformer enhances the translational efficiency of viral RNAs in vivo 40-fold and blocks viral minus-strand RNA synthesis in vitro. AMV CP is proposed to initiate infection by mimicking the function of the poly(A)-binding protein.

Cucumoviruses

Research on the molecular biology of cucumoviruses and their plant-virus interactions has been very extensive in the last decade. Cucumovirus genome structures have been analyzed, giving new insights into their genetic variability, evolution, and taxonomy. A new viral gene has been discovered, and its role in promoting virus infection has been delineated. The localization and various functions of each viral-encoded gene product have been established. The particle structures of Cucumber mosaic virus (CMV) and Tomato aspermy virus have been determined. Pathogenicity domains have been mapped, and barriers to virus infection have been localized. The movement pathways of the viruses in some hosts have been discerned, and viral mutants affecting the movement processes have been identified. Host responses to viral infection have been characterized, both temporally and spatially. Progress has been made in determining the mechanisms of replication, gene expression, and transmission of CMV. The pathogenicity determinants of various satellite RNAs have been characterized, and the importance of secondary structure in satellite RNA-mediated interactions has been recognized. Novel plant genes specifying resistance to infection by CMV have been identified. In some cases, these genes have been mapped, and one resistance gene to CMV has been isolated and characterized. Pathogen-derived resistance has been demonstrated against CMV using various segments of the CMV genome, and the mechanisms of some of these forms of resistances have been analyzed. Finally, the nature of synergistic interactions between CMV and other viruses has been characterized. This review highlights these various achievements in the context of the previous work on the biology of cucumoviruses and their interactions with plants.

Autonomous role of 3'-terminal CCCA in directing transcription of RNAs by Qbeta replicase

We have studied transcription in vitro by Qbeta replicase to deduce the minimal features needed for efficient end-to-end copying of an RNA template. Our studies have used templates ca. 30 nucleotides long that are expected to be free of secondary structure, permitting unambiguous analysis of the role of template sequence in directing transcription. A 3'-terminal CCCA (3'-CCCA) directs transcriptional initiation to opposite the underlined C; the amount of transcription is comparable between RNAs possessing upstream (CCA)(n) tracts, A-rich sequences, or a highly folded domain and is also comparable in single-round transcription assays to transcription of two amplifiable RNAs. Predominant initiation occurs within the 3'-CCCA initiation box when a wide variety of sequences is present immediately upstream, but CCA or a closely similar sequence in that position results in significant internal initiation. Removal of the 3'-A from the 3'-CCCA results in 5- to 10-fold-lower transcription, emphasizing the importance of the nontemplated addition of 3'-A by Qbeta replicase during termination. In considering whether 3'-CCCA could provide sufficient specificity for viral transcription, and consequently amplification, in vivo, we note that tRNA(His) is the only stable Escherichia coli RNA with 3'-CCCA. In vitro-generated transcripts corresponding to tRNA(His) served as poor templates for Qbeta replicase; this was shown to be due to the inaccessibility of the partially base-paired CCCA. These studies demonstrate that 3'-CCCA plays a major role in the control of transcription by Qbeta replicase and that the abundant RNAs present in the host cell should not be efficient templates.

Characterization of the cis-acting elements controlling subgenomic mRNAs of citrus tristeza virus: production of positive- and negative-stranded 3'-terminal and positive-stranded 5'-terminal RNAs

Citrus tristeza virus (CTV), a member of the Closteroviridae, has an approximately 20-kb positive-sense RNA genome with two 5' ORFs translated from the genomic RNA and 10 3' genes expressed via nine or ten 3'-terminal subgenomic (sg) RNAs. The expression of the 3' genes appears to have properties intermediate between the smaller viruses of the "alphavirus supergroup" and the larger viruses of the Coronaviridae. The sgRNAs are contiguous with the genome, without a common 5' leader, and are associated with large amounts of complementary sgRNAs. Production of the different sgRNAs is regulated temporally and quantitatively, with the highly expressed genes having noncoding regions (NCR) 5' of the ORFs. The cis-acting elements that control the highly expressed major coat protein (CP) gene and the intermediately expressed minor coat protein (CPm) gene were mapped and compared. Mutational analysis showed that the CP sgRNA controller element mapped within nts -47 to -5 upstream of the transcription start site, entirely within the NCR, while the CPm control region mapped within a 57 nt sequence within the upstream ORF. Although both regions were predicted to fold into two stem-loop structures, mutagenesis suggested that primary structure might be more important than the secondary structure. Because each controller element produced large amounts of 3'-terminal positive- and negative-stranded sgRNAs, we could not differentiate whether the cis-acting element functioned as a promoter or terminator, or both. Reversal of the control element unexpectedly produced large amounts of a negative-stranded sgRNA apparently by termination of negative-stranded genomic RNA synthesis. Further examination of controller elements in their native orientation showed normal production of abundant amounts of positive-stranded sgRNAs extending to near the 5'-terminus, corresponding to termination at each controller element. Thus, each controller element produced three sgRNAs, a 5'-terminal positive strand and both positive- and negative-stranded 3'-terminal RNAs. Therefore, theoretically CTV could produce 30-33 species of RNAs in infected cells.

Soluble, template-dependent extracts from Nicotiana benthamiana plants infected with potato virus X transcribe both plus- and minus-strand RNA templates

We have developed a method to convert membrane-bound replication complexes isolated from Nicotiana benthamiana plants infected with potato virus X (PVX) to a soluble, template-dependent system for analysis of RNA synthesis. Analysis of RNA-dependent RNA polymerase activity in the membrane-bound, endogenous template extracts indicated three major products, which corresponded to double-stranded versions of PVX genomic RNA and the two predominant subgenomic RNAs. The endogenous templates were removed from the membrane-bound complex by treatment with BAL 31 nuclease in the presence of Nonidet P-40 (NP-40). Upon the addition of full-length plus- or minus- strand PVX transcripts, the corresponding-size products were detected. Synthesis was not observed when red clover necrotic mosaic dianthovirus (RCNMV) RNA 2 templates were added, indicating template specificity for PVX transcripts. Plus-strand PVX templates lacking the 3' terminal region were not copied, suggesting that elements in the 3' region were required for initiation of RNA synthesis. Extracts that supported RNA synthesis from endogenous templates could also be solublized using sodium taurodeoxycholate and then rendered template-dependent by BAL 31 nuclease/NP-40 treatment. The solubilized preparations copied both plus- and minus-strand PVX transcripts, but did not support synthesis from RCNMV RNA 2. These membrane-bound and soluble template-dependent systems will facilitate analyses of viral and host components required for PVX RNA synthesis.

Initiation of genomic plus-strand RNA synthesis from DNA and RNA templates by a viral RNA-dependent RNA polymerase

In contrast to the synthesis of minus-strand genomic and plus-strand subgenomic RNAs, the requirements for brome mosaic virus (BMV) genomic plus-strand RNA synthesis in vitro have not been previously reported. Therefore, little is known about the biochemical requirements for directing genomic plus-strand synthesis. Using DNA templates to characterize the requirements for RNA-dependent RNA polymerase template recognition, we found that initiation from the 3' end of a template requires one nucleotide 3' of the initiation nucleotide. The addition of a nontemplated nucleotide at the 3' end of minus-strand BMV RNAs led to initiation of genomic plus-strand RNA in vitro. Genomic plus-strand initiation was specific since cucumber mosaic virus minus-strand RNA templates were unable to direct efficient synthesis under the same conditions. In addition, mutational analysis of the minus-strand template revealed that the -1 nontemplated nucleotide, along with the +1 cytidylate and +2 adenylate, is important for RNA-dependent RNA polymerase interaction. Furthermore, genomic plus-strand RNA synthesis is affected by sequences 5' of the initiation site.

Infectious RNA transcripts from grapevine virus A cDNA clone

A full length cDNA clone of grapevine virus A (GVA) was constructed downstream from the bacteriophage T7 RNA polymerase promoter. Capped in vitro-transcribed RNA was infectious in Nicotiana benthamiana and N. clevelandii plants. Symptoms induced by the RNA transcripts or by the parental virus were indistinguishable. The infectivity of the in vitro-transcribed RNA was confirmed by serological detection of the virus coat and movement proteins and by observation of virions by electron microscopy. This is the first report of infectious RNA transcripts derived from a full-length cDNA clone of a member of the Vitivirus genus.

Genes required for replication of the 15.5-kilobase RNA genome of a plant closterovirus

A full-length cDNA clone of beet yellows closterovirus (BYV) was engineered and used to map functions involved in the replication of the viral RNA genome and subgenomic RNA formation. Among 10 open reading frames (ORFs) present in BYV, ORFs 1a and 1b suffice for RNA replication and transcription. The proteins encoded in these ORFs harbor putative methyltransferase, RNA helicase, and RNA polymerase domains common to Sindbis virus-like viruses and a large interdomain region that is unique to closteroviruses. The papain-like leader proteinase (L-Pro) encoded in the 5'-proximal region of ORF 1a was found to have a dual function in genome amplification. First, the autocatalytic cleavage between L-Pro and the remainder of the ORF 1a product was essential for replication of RNA. Second, an additional L-Pro function that was separable from proteolytic activity was required for efficient RNA accumulation. The deletion of a large, approximately 5.6-kb, 3'-terminal region coding for a 6-kDa hydrophobic protein, an HSP70 homolog, a 64-kDa protein, minor and major capsid proteins, a 20-kDa protein, and a 21-kDa protein (p21) resulted in replication-competent RNA. However, examination of mutants with replacements of start codons in each of these seven 3'-terminal ORFs revealed that p21 functions as an enhancer of genome amplification. The intriguing analogies between the genome organization and replicational requirements of plant closteroviruses and animal coronavirus-like viruses are discussed.

Template-independent repair of the 3' end of cucumber mosaic virus satellite RNA controlled by RNAs 1 and 2 of helper virus

RNA viruses which do not have a poly(A) tail or a tRNA-like structure for the protection of their vulnerable 3' termini may have developed a different strategy to maintain their genome integrity. We provide evidence that deletions of up to 7 nucleotides from the 3' terminus of cucumber mosaic cucumovirus (CMV) satellite RNA (satRNA) were repaired in planta in the presence of the helper virus (HV) CMV. Sequence comparison of 3'-end-repaired satRNA progenies, and of satRNA and HV RNA, suggested that the repair was not dependent on a viral template. The 3' end of CMV satRNA lacking the last three cytosines was not repaired in planta in the presence of tomato aspermy cucumovirus (TAV), although TAV is an efficient helper for the replication of CMV satRNA. With use of pseudorecombinants constructed by the interchange of RNAs 1 and 2 of TAV and CMV, evidence was provided that the 3'-end repair was controlled by RNAs 1 and 2 of CMV, which encode subunits of the viral RNA replicase. These results, and the observation of short repeated sequences close to the 3' terminus of repaired molecules, suggest that the HV replicase maintains the integrity of the satRNA genome, playing a role analogous to that of cellular telomerases.

Comparison of the replication of positive-stranded RNA viruses of plants and animals

It is clear from the experimental data that there are some similarities in RNA replication for all eukaryotic positive-stranded RNA viruses—that is, the mechanism of polymerization of the nucleotides is probably similar for all. It is noteworthy that all mechanisms appear to utilize host membranes as a site of replication. Membranes appear to function not only as a way of compartmentalizing virus RNA replication but also appear to have a central role in the organization and functioning of the replication complex, and further studies in this area are needed. Within virus supergroups, similarities are evident between animal and plant viruses—for example, in the nature and arrangements of replication genes and in sequence similarities of functional domains. However, it is also clear that there has been considerable divergence, even within supergroups. For example, the animal alpha-viruses have evolved to encode proteinases which play a central controlling function in the replication cycle, whereas this is not common in the plant alpha-like viruses and even when it occurs, as in the tymoviruses, the strategies that have evolved appear to be significantly different. Some of the divergence could be host-dependent and the increasing interest in the role of host proteins in replication should be fruitful in revealing how different systems have evolved. Finally, there are virus supergroups that appear to have no close relatives between animals and plants, such as the animal coronavirus-like supergroup and the plant carmo-like supergroup.

Synthesis of novel products in vitro by an RNA-dependent RNA polymerase

RNA-dependent RNA polymerase from turnip crinkle virus-infected turnip transcribes both strands of a virus-associated satellite RNA, sat-RNA C (356 bases), in vitro. While both plus- and minus-strand sat-RNA C can direct the synthesis of full-length complementary-strand products, transcription of minus-strand RNA also generates two non-template-sized products, L-RNA and S-RNA (C. Song and A. E. Simon, Proc. Natl. Acad. Sci. USA 91:8792-8796, 1994). Here we report that synthesis of L-RNA and S-RNA results from terminal elongation of the 3' end of the template. L-RNA has a panhandle structure and is composed of minus-strand template covalently linked to newly synthesized RNA complementary to its 5' 190 bases. S-RNA is composed of template covalently linked to its full-length complementary strand. All minus-strand templates tested yielded S-RNA. However, synthesis of L-RNA was affected by deletion of the 3' end of the minus-strand template or several internal regions and base alterations near the 5' end or in an internal sequence immediately upstream from the template-product junction that could potentially form a heteroduplex with the 3' end. Furthermore, mutations that disrupted or restored a stem-loop involved in RNA recombination in vivo affected the level of L-RNA produced in vitro, suggesting that the mechanisms for intramolecular formation of panhandle RNAs and intermolecular RNA recombination involve similar features.

Bromovirus RNA replication and transcription require compatibility between the polymerase- and helicase-like viral RNA synthesis proteins

The positive-strand RNA bromoviruses encode two nonstructural proteins, 1a and 2a, involved in RNA-dependent RNA replication. These proteins have extensive sequence similarities with methyltransferase, helicase, and polymerase proteins of other plant and animal viruses. 1a and 2a can also form a complex in vitro. To explore whether 1a-2a interaction is required for RNA replication in vivo, we reassorted the 1a and 2a genes from two different bromoviruses, brome mosaic virus (BMV) and cowpea chlorotic mottle virus (CCMV). 1a and 2a were expressed independently of viral replication by using RNA- or DNA-based transient expression, and their in vivo RNA replication activities were tested in protoplasts with BMV and CCMV RNA3 templates. RNA-based transient expression confirmed prior indications that bromovirus RNA replication is more sensitive to reductions in 1a expression than to reductions in 2a expression. DNA-based expression of the homologous combinations of 1a and 2a supported high levels of RNA synthesis, but both 1a-2a heterologous combinations exhibited RNA synthesis defects. The combination of CCMV 1a and BMV 2a did not support detectable synthesis of negative-strand, positive-strand, or subgenomic RNA. The converse combination of BMV 1a and CCMV 2a was preferentially defective in positive-strand and subgenomic RNA accumulation, showing that 1a-2a interaction is involved in these processes in ways distinct from negative-strand RNA synthesis, which was only slightly affected. These results indicate that at least some functions of 1a and 2a operate in a mutually dependent manner in vivo and that the mechanisms of positive- and negative-strand RNA synthesis are differentiated in part by features of such interactions.

Formation of multimers of linear satellite RNAs

A 22-base region of turnip crinkle virus satellite-RNA C (sat-RNA C) is involved in the accumulation of monomeric and dimeric forms. Deletions within the region inhibited the accumulation of sat-RNA C monomers. However, normal ratios of dimers to monomers occurred if the 22 bases were replaced by 22 unrelated bases or if the location of this region was altered. Therefore, these specific 22 bases are not involved in the accumulation of sat-RNA C monomers. Examination of the sequences at the junctions of multimers of all three turnip crinkle virus sat-RNAs revealed the deletion of bases corresponding to the 3' and 5' ends of monomeric units as well as the addition of nucleotides not present in monomers. Based on these results, we present a model to explain the formation of multimers of linear subviral RNAs associated with turnip crinkle virus. Our model suggests that multimers are formed by the reinitiation of replication by the replicase before release of the nascent strand. We have previously proposed the same mechanism for the formation of defective interfering RNAs, chimeric sat-RNAs, and sat-RNA recombinants in the turnip crinkle virus system (Cascone, Carpenter, Li, and Simon. (1990). EMBO J. 9, 1709-1715).

Similarities among plant virus (+) and (-) RNA termini imply a common ancestry with promoters of eukaryotic tRNAs

The 5' ends of brome mosaic virus (BMV) RNAs contain sequences similar to the consensus internal control region (ICR) of pol III promoters in tRNA genes. Comparison of BMV (+)RNA 5' termini with BMV (-)RNA termini revealed the presence of two (tandem) repeats of some 30 nucleotides, the more internal containing a region of 73% similarity to the tRNA consensus ICR2 (downstream) region of the ICR. Tandem repeats containing motifs similar to the ICR2 consensus were found at the 5' termini of (-)RNAs of cucumo-, tobamo-, and tymoviruses whose 3' (+)RNAs have aminoacylatable tRNA-like structures. Single regions of homology to the BMV(+)RNA 5' terminus, containing an ICR2-like motif, were detected for several tobravirus RNAs, and for satellite tobacco necrosis virus RNA. The (+)-stranded genomes of these viruses have not been shown to be capable of amino acid esterification. The ICR2 consensus (GGUUCGANUCC) is nearly palindromic, and is contained with the T psi C loop of tRNAs and viral analogs. Consequently, tRNA promoter-like motifs can be seen at both termini of (+) and (-) RNAs of bromoviruses and other viruses. The presence of ICR1 and ICR2-like sequences in BMV genomic 5' (+)RNAs and the tobamovirus 5' (-)RNAs may reflect promoter arrangements of primordial genomic RNAs ancestral to both modern plant viruses and eukaryotic tRNAs. Several derivative concepts related to genome evolution are discussed, including the origin of asymmetric strand synthesis of RNAs.

Self-cleavage of plus and minus RNAs of a virusoid and a structural model for the active sites

Virusoids are circular single-stranded RNAs dependent on plant viruses for replication and encapsidation. Virusoid replication appears to involve longer-than-unit-length plus and minus RNAs, indicating that unit-length plus RNA is generated by specific cleavage reactions. Here, we synthesize plus and minus partial-length RNAs of the 324-nucleotide virusoid from lucerne transient streak virus in vitro. Both RNAs self-cleave at a unique site in the presence of magnesium ions to give 5' hydroxyl and 2',3' cyclic phosphodiester termini. Conformations other than the native structures are necessary for cleavage. Similar secondary structures with considerable sequence homology are proposed for the active sites of these and other plant pathogenic RNAs. Our results are consistent with certain rolling-circle replication models.

Isolation and partial characterization of a monoclonal antibody specific for a naturally occurring double-stranded RNA

A monoclonal antibody (Mab) against the double-stranded form of the cucumber mosaic virus-associated RNA 5, the sequence variant that induces the tomato necrosis disease (ds(n)CARNA 5), has been obtained. Specificity studies using a radioimmunoassay showed that the purified Mab recognizes neither DNA nor ssRNA of different sources including ss(n)CARNA 5, poly(I).poly(C), nor dsRNA of two other sequence variants of CARNA 5. Therefore, the Mab seems to recognize a conformational variation only present in the double helical structure of ds(n)CARNA 5.

Minus-strand initiation by brome mosaic virus replicase within the 3' tRNA-like structure of native and modified RNA templates

An RNA-dependent RNA polymerase (replicase) extract from brome mosaic virus-infected barley leaves has been shown to initiate synthesis of (-) sense RNA from (+) sense virion RNA. Initiation occurred de novo, as demonstrated by the incorporation of [gamma-32P]GTP into the product. Sequencing using cordycepin triphosphate to terminate (-) strands during their synthesis by the replicase generated sequence ladders that confirmed that copying was accurate, and that initiation occurred very close to the 3' end. The precise site of initiation was further defined by testing the replicase template activity after stepwise removal of 3'-terminal nucleotides. Whereas removal of the terminal A did not decrease template activity, removal of the next nucleotide (C-2) did. Thus, initiation almost certainly occurs opposite the penultimate 3'-nucleotide (C-2) in vitro. The structure of the double-stranded replicative form of RNA isolated from brome mosaic virus-infected leaves was consistent with such a mechanism occurring in vivo, in that it lacked the 3'-terminal A found on virion RNAs. The specific site of (-) strand initiation and normal template activity were retained for RNAs with as many as 15 to 30 A residues added to the 3' end. However, only limited oligonucleotide 3' extensions can be present on active templates. In order to assess the 5' extent of sequences required for an active template, a 134-nucleotide-long fragment of brome mosaic virus RNA, corresponding to the tRNA-like structure, was generated. This RNA had high template activity, but a shorter 3' (85-nucleotide) fragment was inactive. RNAs with various heterologous sequences 5' to position 134 also showed high template activity. Thus, the 3'-terminal tRNA-like structure common to all four brome mosaic virus virion RNAs contains all of the signals required for initiation of replication, and sequences 5' to it do not play a role in template selection.