Synthesis of 3'-coterminal sgRNAs is one of the genome expression strategies adapted by members of the family Closteroviridae. Evidence gathered with viruses in the genera Closterovirus [4, 6] and Crinivirus [5, 18, 35] demonstrated that the temporal expression and kinetics of accumulation of these sgRNAs is highly regulated and level of sgRNAs expressed depends on promoter strength and position within the genome. However, expression strategies of sgRNAs for members of the genus Ampelovirus have not been studied to date, despite their economic importance to many agriculturally important crops. The characteristic profiles of sgRNAs obtained in Northern blots from virus-infected grapevine (Fig. 2) provided evidence that GLRaV-3 likely employs a strategy for the expression of its nested set of 3'-coterminal sgRNAs similar to other closteroviruses. However, accumulation of sgRNAs detected in virus-infected grapevine tissue indicated differences in the levels of expression of specific sgRNAs compared to other monopartite closteroviruses like BYV and CTV [15, 16, 36, 37]. Since results of this study were from an asynchronous infection, it is possible that the amounts and timing of synthesis of sgRNAs in GLRaV-3-infected tissue are variable during the growing season and remains to be elucidated to determine whether their profile changes in relation to the developmental stage or expression of grapevine leafroll disease symptoms .
The sgRNAs for p7 and p4 ORFs, located close to the 3' terminus of GLRaV-3 genome, accumulated at lower levels than sgRNAs specific to upstream ORFs p20B, p20A and p21. Similarly, the sgRNA specific to the CP showed higher accumulation than the sgRNA specific to CPm, which is closer to the 3' terminus. Another unique feature of GLRaV-3 is that the sgRNAs specific to p21 ORF, which has only a 10 nt 5'nontranslated region upstream of the ORF where a cis-acting promoter element would be expected to occur, and the sgRNA corresponding to the p20A ORF, which lacks an upstream nontranslated region due to overlap with the upstream p21 ORF, accumulate at more or less similar levels compared to sgRNA of the CP ORF that has an 89 nt upstream nontranslated region. These results clearly suggest that GLRaV-3 does not follow either of the two general rules: (i) genes located nearer the 3'terminus are usually expressed at higher levels and (ii) ORFs with a 5'nontranslated region are generally expressed higher than ORFs that overlap with the upstream ORF (and hence no nontranslated region). Instead, it is likely that production of GLRaV-3 sgRNAs follows an alternative or modified mechanism.
The mapping of 5' termini of four 3'-terminal genes (CP, p21, p20A and p20B) of GLRaV-3 indicated that they possess a +1 adenylate, same as the 5' end of the genome, and this observation was analogous to CTV than to other closteroviruses like BYV [16–18]. The leader sequences of GLRaV-3 sgRNAs are collinear with the genomic RNA and their relative lengths are within the range observed for sgRNAs of CTV and BYV. The first few nucleotides of the 5' end of the genomic RNA (ATAAATG) and the genomic sequence around the 5' termini of sgRNAs (underlined, CP [TTTTATAAA], p21 [TCTTAGAAA], p20A [ATTGAATGA], p20B [AAGTATATT]) were similar but not identical AU-rich regions. Presence of an adenylate as the 5' terminus in the genomic and all four sgRNAs suggested that initiation of RNA synthesis from an uridylate on the minus strand is preferred by the GLRaV-3 replicase complex, similar to that proposed for CTV . However, the lack of sequence conservation around the 5' termini and the absence of secondary structure elements or octanucleotides conserved in BYV and CTV nontranslated leader regions suggest that sgRNA expression strategy in GLRaV-3 could be somewhat different leading to the hypothesis that some commonalities do exist in the replication strategy between closteroviruses and ampeloviruses, but also some features characteristic to ampeloviruses.
In this study, the complete genomic RNA sequence of GLRaV-3 was determined to consist of 18,498 nt and is in agreement with the size of GLRaV-3 isolate reported from South Africa . The overall genome organization of these two isolates is similar to that of GLRaV-3 isolates sequenced from New York and Chile. All four isolates have highly conserved 3'NTRs than 5'NTRs due to the unusual variation in size and sequence between them. Given the same size 5'NTR in two GLRaV-3 genome sequences obtained by independent groups in distant locations [this study and 23], it is likely that artifacts in cloning would have resulted in the apparent small size 5'NTR reported for New York and Chile isolates. It could be possible that use of poly (A) tailing to amplify the 5'NTR and the likelihood of non-specific annealing of oligo (dT) primer to a portion of 5'NTR with high adenines in the complementary DNA strand of GLRaV-3 would have contributed to the amplification of less-than full size sequence at the 5' terminal portion of the genome . The 5'portion of GLRaV-3 isolate from Chile was amplified by RT-PCR using primers designed based on the sequence of New York isolate  and hence may not be a true representation of the authentic 5'end of virus genome. In contrast, our study and that reported from South Africa  used 5'RACE to accurately map the 5'end of GLRaV-3 genome. Additionally, same size PCR products with high level of sequence similarity amplified from GLRaV-3 isolates originating from different grapevine cultivars planted in geographically distinct locations in the US provided additional evidence that the size of 5'NTR determined in our study is accurate.
A 737 nt long 5'NTR in GLRaV-3 genome represents the longest 5'NTR among the currently known monopartite members of the family Closteroviridae available in database. The sizes of 5'NTR in viruses of the genus Closterovirus is between 105 and 227 nt and those in the genus Ampelovirus between 213 and 737 nt, respectively. In contrast, the bipartite and tripartite members of the genus Crinivirus have 72 to 264 nt 5'NTR in RNA 1. Thus, highly variable size of 5'NTR appears to be a characteristic of the members in the genus Ampelovirus in the family Closteroviridae. In addition, the 5'NTRs of grapevine-infecting closteroviruses, such as GLRaV-1, GLRaV-2 and GLRaV-Pr, for which complete genome sequences are available in the database, are variable in size with no significant sequence homology between them. Outside of the family Closteroviridae, a long 5'NTR of 739 nt was recently observed in Triticum mosaic virus (TriMV), a wheat-infecting virus in the family Potyviridae . To our knowledge, GLRaV-3 and TriMV appears to be the only plant viruses with monopartite genome known to date with such a long sequence in their 5'NTR. A long 5'NTR in plant viruses is unusual, but a long 5'NTR ranging in length from 610 to 1500 nt has been reported in animal/human-infecting viruses in the family Picornaviridae . Unlike 5'NTRs of TriMV and picornaviruses, which contain multiple non-conserved AUG upstream of the initiation codon, the 737 nt long 5'NTR of GLRaV-3 isolates contain only one AUG triplet at the very 5'terminus (5'-ATAAATGCTC) preceding the translation initiation codon of the ORF1 polyprotein. Thus, the long 5'NTR of GLRaV-3 appears to have features distinct from other plant and animal infecting viruses. The 5'NTR sequences in many picornaviruses and flaviviruses are highly structured to form internal ribosome entry (IRES) elements that play a role in protein translation in a cap-independent manner . It remains to be studied if any potential IRES elements exist in the 5'NTR of GLRaV-3. At a practical level, the high variability in 5'NTR could be useful for discriminating GLRaV-3 isolates from different grape-growing areas around the world into phylogenetically distinct lineages.
In the case of CTV, the 5'NTRs of all isolates could be folded into two similar stem-loop structures despite the fact that their sequence varied by as much as 58% . These two conserved structures were shown to be important for replication and assembly [7, 42]. Even though we showed a complex secondary structure for 5'NTR of GLRaV-3 isolates (Fig. 5a) as predicted by using the MFOLD program, the complex structures did not seem to be conserved between the Washington and South Africa isolates. A tandem repeat of 65 nt in the 5'NTR sequence of many GLRaV-3 isolates from the US, but not in the 5'NTR of the South Africa isolate, is surprising and the functional importance of this repeat sequence in virus life cycle needs to be examined. In contrast to the 5'NTR, the 3'NTR of the four GLRaV-3 isolates is highly conserved and folded into identical secondary structure (Fig. 5b). This pattern is analogous to CTV, but different from BYV, where different isolates were shown to have considerable differences in size and sequence identity in the 3'NTR . It has been shown with CTV that the 3'NTR contains cis-acting elements required to initiate synthesis of complementary negative strands and some of the predicted secondary structures are important components for efficient replication  and it is likely that the predicted SL structures in the 3'NTR contain cis-acting elements to play a critical role in GLRaV-3 replication.