key: cord-0696484-0kcujnfa
authors: Agranovsky, A. A.
title: Principles of Molecular Organization, Expression, and Evolution of Closteroviruses: Over The Barriers
date: 1996-12-31
journal: Advances in Virus Research
DOI: 10.1016/s0065-3527(08)60735-6
sha: 27b2c24e392bcbd8ad16a42aab322ea091cc99e5
doc_id: 696484
cord_uid: 0kcujnfa

Publisher Summary This chapter focuses on the molecular organization, evolution, and expression of closterovirus genomes, as well as on their unique particle structure. The closterovirus group combines several positive-strand RNA viruses with very flexuous filamentous particles, of which beet yellows virus (BYV) is the type virus. Closteroviruses are distinct from other RNA viruses of plants in some important phenomenological aspects. They have genomes of up to 20 kilobases (kb), a value comparable only to those of the animal coronaviruses and toroviruses, which have the largest RNA genomes of all positive-strand RNA viruses. The existence of such genomes having a coding capacity several times that of an average RNA virus genome raises questions as to the trend whereby the long genomes have evolved and the possible novel functions they have acquired. The dramatic increase in the closterovirus genome coding capacity may be linked to the distinct ecological niche they occupy. Thus, closteroviruses are the only elongated plant viruses known so far to cause phloem-limited infections in plants and to persist in their insect vectors for many hours, in contrast to only minutes.

The closterovirus group (having its name from K~U U E~, Greek for "thread") combines several positive-strand RNA viruses with very flexuous filamentous particles, of which beet yellows virus (BYV) is the type virus (Bar-Joseph and Hull, 1974; Bar-Joseph and Murant, 1982) . Closteroviruses are distinct from other RNA viruses of plants in some important phenomenological aspects. They have genomes of up to 20 kilobases (kb), a value comparable only to those of the animal coronaviruses and toroviruses, which have the largest RNA genomes of all positive-strand RNA viruses. The existence of such genomes having a coding capacity several times that of an average RNA virus genome (e.g., TMV) raises questions as to the trend whereby the long genomes 11. DRAFT OF CLOSTEROVIRUS TAXONOMY Until recently, the taxonomic status of closteroviruses has been illdefined, and the group was a quite heterogeneous collection, with the main distinction being between the so-called typical and atypical closteroviruses (Bar-Joseph et al., 1979; Bar-Joseph and Murant, 1982; Francki et al., 1991) . Typical closteroviruses have particles of 1000-2000 nm in length, induce characteristic BYV-type membranaceous vesicles in infected tissues, and are transmitted by insect vectors (Table 1) . Atypical (or "clostero-like") viruses have shorter particles of about 750 nm, do not induce BYV-type vesicles, and have no known vector. Sequencing of the typical and atypical closteroviruses and comparison of the encoded protein sequences confirmed the disparity between them (Agranovsky et al., 1991a (Agranovsky et al., ,b, 1994a Dolja et ul., 1991 Dolja et ul., , 1994 German ~t al., 1990; Koonin, 1991; Yoshikawa et al., 1992) . Former closteroviruses such as apple chlorotic leafspot virus, apple stem grooving virus and citrus tatter leaf virus have already been reclassified as trichoviruses and capilloviruses, respectively (Francki et al., 1991;  Gunasinghe and German (1989) Yes Lockhart et al. (1992) Yes Inouye (1976) Yes Yes Duffus et al. (1986) ; Klaasen et al.

(1995) Cohen et al. (1992) ; Winter et al. (1992) ; U. Hoyer, personal communication Some viruses currently recognized as closteroviruses, i.e., lettuce chlorosis virus (LeCV), tomato infectious chlorosis virus (TICV), cucurbit yellow stunting disorder virus (CYSDV), and grapevine leafroll-associated viruses 5 and 6 (GLRaV-5 and -6), are not included because of the lack of descriptive data; on the other hand, LCV, which is not yet considered a family member by ICTV, is included here since its features appear to be typical of the Closterouiridae.

The exact molecular weight values are indicated wherever sequence data were available.

Abbreviations used for vectors are A, aphids; M, mealybugs; W, whiteflies.

Tentative species.

' NK, not known.

' B P W has several synonymic designations, i.e., cucumber yellows, muskmelon yellows, or melon yellows virus (J. P. Martelli, personal communication); the question of whether B P W and cucumber chlorotic spot virus (CCSV; Section V,B) are identical is uncertain and thus awaits comparative tests (L. P. Woudt, personal communication). As indicated by recent analysis (Boscia et al., 1995) , GLRV-2 is identical to the virus earlier described as "grapevine corky barkassociated virus" (GCBaV; Namba et al., 1991) .

L I W has particles of 950 nm modal length according to other measurements (D.-E. Lesemann, personal communication) . 124 A. A. AGRANOVSKY (SPSW, Cohen et al., 1992) , and beet pseudo-yellows virus (BPYV; Duffus, 1973) . A semipersistent transmission pattern implies a stricter virus-vector specificity as compared with, for example, the nonpersistent aphid transmission of potyviruses (Falk and Duffus, 1988) . Indeed, among seven aphid species that can transmit CTV, Toxoptera citridicis is the most efficient vector; likewise, Myzus persicae is the best among 24 species transmitting B W (Duffus, 1973; Bar-Joseph et al., 1979; Lister and Bar-Joseph, 1981) . Interactions of the whiteflyborne closteroviruses with their vectors seem to be even more specific; thus, L I W and S P S W can only be transmitted by Bemicia tabaci (Cohen et al., 1992; Iluffus et al., 1986) , whereas B P W can only be transmitted by Dialeurodes uaporariorum (Duffus, 1973) . Likewise, the mealybug-borne grapevine leafroll-associated virus 3 (GLRaV-3) and the little cherry virus (LCV) are specifically transmitted by Puluinaria uitis and Phenacoccus aceris, respectively (Belli et al., 1994; Raine el al., 1986) . The time needed for aphids to successfully inoculate closteroviruses and the type of the disease symptoms may both reflect the phloemlimited nature of closteroviral infections (Bar-Joseph et al., 1979; Milne, 1988) . Data obtained with the help of electronic monitoring indicate that, on B W transmission, the aphid stylet reaches the phloem in 9-38 min (Limburg et al., 1994) .

In the cells, closteroviruses give rise to massive particle aggregates often organized as banded inclusions. The most characteristic type of intracellular inclusions are numerous vesicles surrounded by a membrane, possibly of mitochondria1 origin (Essau and Hoefert, 1971; Lesemann, 1988; Kim et al., 1989) . These structures, referred to as BYV-type vesicles, are considered an important taxonomic feature of the family, and are suitable for closterovirus diagnosis (Table I) . The BYV-type vesicles presumably contain double-stranded RNA (Coffin and Coutts, 1993) ; it would be interesting to determine whether these structures have anything to do with closterovirus replication. (Table I) , and about 12 nm in diameter (Tollin and Wilson, 1988 CTV have a primary helix pitch of 3.5-3.7 nm, with 8.5 and 10 subunits of CP per helix turn, respectively (Bar-Joseph et al., 1972; Chevallier et al., 1983) . In the B W particle, the structure repeats in two helical turns (Chevallier et al., 1983) , in contrast to the heracleum virus 6 structure with a five-turns repeat (Tollin et al., 1992) . The fact that closteroviruses possess the most flexible particles among elongated RNA viruses has been a key feature for their recognition as a distinct group (Brandes and Bercks, 1965) . This apparent flexibility is reflected in a lower ratio of closterovirus RNA mass to modal particle length (2831-3230hm) as compared with that for potex-, poty-, and carlaviruses (4038--4112/nm), or tobamoviruses (6666hm) (Bar-Joseph and Hull, 1974; Bar-Joseph et al., 1979) . A loosely wound helix of B W and CTV particles may account for their sensitivity to ribonuclease (Bar-Joseph and Hull, 1974) .

It has been found that the B W genome encodes a 24-kDa protein (24K) strikingly similar to the 22-kDa CP (22K), and that both these proteins have counterparts encoded in the CTV genome (Boyko et al., 1992) . It has been suggested that the genes for CP homologues arose by gene duplication that probably occurred in a common closterovirus ancestor; it is noteworthy that, despite significant divergence, the CP homologues of BYV and CTV have retained the profile of conserved amino acid residues that are believed to ensure the characteristic fold of the filamentous plant virus CPs (Boyko et al., 1992 , Dolja et al., 1991 . This discovery posed a question of the presence of a second structural protein in the virions. Initially, there were doubts as to whether this is theoretically possible, and researchers in this field (the author being no exception) have long been under the spell of the "single C P paradigm of elongated plant virus structure (Boyko et al., 1992; Dolja et al., 1994) . Then, quite unexpectedly, immunoelectron microscopy (IEM) revealed two serologically distinct segments in the B W particles: a "tail" selectively labeled with antibodies to the 24-kDa protein (Fig. lA) , and the main segment labeled with antibodies to purified virus and to the recombinant 22-kDa protein ( Fig. 1B ; D.-E. Lesemann and author, unpublished observations) . After decoration with antibodies to the 24-kDa protein, the B W particles exhibited two peaks of modal lengths corresponding to tailed 1370-nm particles and to 1293-nm tailless particles; the lengths of the anti-24K serumdecorated tails peaked at 75 nm . The good balance between these lengths illustrates the propensity of the tails to break off and the relative stability of the 22K-coated part against further degradation. In line with this, purified B W preparations contained a lower portion of the tailed particles as compared with crude A B PIG. 1. Iminunoelectron micrographs of BYV part,icles in crude sap of infected Tetragoizia expanna plants (~95,000 magnification). (A) Decoration with mouse polyclonal antiserum against the N-terminal peptide of the BYV 24K protein. (B) Decoration wit,h rabbit antiserum against purified BYV particlcs. Grids with the adsorbed virus particles were incubated with the antiscra for 15 min (A) and 30 min (B), followed by negative staining with uranyl acetate. Arrows tag the distinct vinon tail. (Courtesy of D.-E. Lesemann.) sap extracts, and the tails were no longer observed in the particles in the infected plant sap after overnight incubation at room temperature (D.-E. Lesemann, unpublished observations) . The discrepancy between the B W particle lengths determined for purified virus preparations and for leaf-dips (1250 versus 1370-1450 nm; reviewed in Bar-Joseph et al., 1979) may also be at least partially attributed to preservation or loss of the tails.

Strange as it may seem, the morphological polarity of the BYV particles was first noticed much earlier. With the aid of electron microscopy of the methylamine tungstate-stained B W particles, Hills and Gay (1976) observed an 83-nm terminal region with a helix pitch of 4.05 nm, which was clearly distinct from the main part of the 1390-nm virion having a pitch of 3.45 nm. These observations, which long remained unexplained, are consistent with the IEM data discussed above.

Thus, B W particles, unlike those of other elongated plant viruses, possess a distinct tail built of multiple subunits of a minor CP, and hence have a "rattlesnake" rather than uniform structure . Comparison of the lengths of the 24K-and 22K-encapsidated segments (75 vs. 1293 nm) gives a rough estimate of one 24-kDa molecule per 17 molecules of 22K, which is consistent with the relative proportions of the respective subgenomic mRNAs in BW-infected tissues (Dolja et al., 1990) . Moreover, the putative subgenomic promoters for the 22K and 24K mRNAs are similar (see Section VI,C), thus suggesting concerted expression of both structural proteins in viral infection. The involvement of 24K in virion formation is in line with previous computer predictions that the closterovirus CP duplicates have a spatial fold conserved in the monophyletic family of filamentous virus CPs (Boyko et al., 1992; Dolja et al., 1991) . Specific decoration of the B W particles with antibodies against the N-terminal peptide of 24K strongly indicates that the N terminus of the minor CP is exposed on the virion surface, as is the case with CPs of filamentous potex-and potyviruses Koenig and Torrance, 1986; Shukla et al., 1988) .

It is possible that other closteroviruses have a similar virion structure. There is remarkable CP size heterogeneity in purified CTV preparations, albeit at least partly due to posttranslational modification of the (major) CP (Sekiya et al., 1991; Lee et al., 1988) . Intriguingly, L I W preparations purified in CszSO4-sucrose gradients contained minor amounts of an approximately 55-kDa protein along with the 28-kDa CP (Klaassen et al., 1994) . The putative CP duplicate encoded in the RNA-2 of this virus has a deduced molecular weight of 52 kDa ( Fig. 2 How is the "rattlesnake" particle assembled? The principal mechanism might follow the classical scheme of TMV self-assembly, which starts at an internal origin of assembly (OAs) in the genomic RNA, and continues by adding CP disks or smaller aggregates in both the 5' and 3' directions (reviewed in Dobrov and Butler, 1984; Mathews, 1991; Lomonossoff and Wilson, 1985) . We have assumed the existence of a nucleation region inherent in the B W RNA that might discriminate between the capsid proteins or their disks, whereupon they proceed to encapsidate the RNA . In order to identify the B W particle's end made up of 24K, we have run the following experiment: purified particles were sonicated and treated with antibodies to 24K, and the antibody-virion fragment complexes were isolated on Protein A-Sepharose. The RNA extracted from the virion fragments thus selected was 5'-end labeled with [Y-~~P]ATP and used as a probe to develop Southern blots with cDNA clones representing the 5'-terminal, middle, and 3'-terminal regions of the B W genome. Compared with the RNA probe prepared from nonfractionated virion fragments, the antibody-selected probe apparently hybridized more strongly to the 5'-terminal clones, thus suggesting that this was the 5'-terminal region of the B W RNA associated with 24K, and that the putative nucleation signal might be mapped to this region (unpublished data) . This is consistent with the fact that none of the 3'coterminal subgenomic RNAs of B W produced in infected plants are found in purified virus preparations (Dolja et al., 1990) . Closterovirus particle formation may be assisted by virus-encoded nonstructural protein(s); in some filamentous DNA phages carrying a few copies of minor CPs at their ends, assembly is chaperoned by phage-encoded proteins that are not part of the mature particles (reviewed in Russel, 1993) . B W encodes at least one nonstructural protein likely to be instrumental in protein-protein interactions, a 65-kDa homologue of the HSP70 cell heat-shock proteins (see Section V,A).

The genome of the Ukrainian strain of B W (BW-U) consists of 15,480 nucleotides (nt), is 5'-capped and contains no 3'-poly(A) (Karasev et al., 1989; Agranovsky et al., 1991b Agranovsky et al., , 1994a . Computer translation reveals nine ORFs in the sequence, flanked by 5'-and 3'-untranslated regions of 107 and 141 nt, respectively (Fig. 2) . For the 3' region, two potentially stable hairpins were predicted, and it has been speculated that these may serve as a recognition signal for viral replicase. B W RNA cannot be aminoacylated or adenylylated in uitro and thus apparently has no 3'-tRNA-like structure (Agranovsky et al., 1991a) .

The 5'-proximal ORFla codes for the 295-kDa product which encompasses the domains with methyltransferase and RNA helicase sequence motifs (MT and HEL; Fig. 2 ) that are conserved in the large subsets of positive-strand RNA viruses (Gorbalenya and Koonin, 1993; Rozanov et al., 1992) . The MT domain is believed to be involved in the capping of viral mRNAs. This activity has been experimentally demon-130 A. A. AGRANOVSKY strated for the alphavirus nsPl protein (Mi and Stollar, 1991) and TMV 126-kDa protein (Dunigan and Zaitlin, 1990) , and suggested for the closely related domains in the Sindbis-like supergroup virus replicases (Ahlquist et al., 1985; Rozanov et al., 1992) . Likewise, the HEL domain, whose strand-separating activity was shown experimentally for the potyvirus CI protein (Lain et al., 1990) and the pestivirus NS3 protein (Warrener and Collett, 1995) , is implicated in unwinding of RNA duplexes on replication of many other virus groups, based on clear conservation of its sequence in the established and putative helicases (Gorbalenya et al., 1988; Koonin, 1989, 1993; Hodgman, 1988) . The ORFlb overlaps the last 40 triplets of ORFla and codes for a product of approximately 53 kDa ( Fig. 2) containing the domains of RNA-dependent RNA polymerase (POL) (Kamer and Argos, 1984; Koonin, 1991) . Putative B W replicase, which is likely to be expressed as an ORFla/lb 348-kDa fusion protein (see below), should have a size about twice that of the other related viral replicases. This difference is due to two unique regions in the putative ORFla/lb fusion product: a 600-residue N-terminal overhang containing a domain of cystein papain-like proteinase (P-Pro), and a 700-residue central insert harboring a 100-residue stretch that may be related to retrovirus aspartyl proteases (Agranovsky ct al., 1994a) . The B W P-Pro was found to be moderately similar to the C-terminal P-Pro domain in helper component proteases (HC-Pro) of potyviruses (Carrington et al., 1989) . HC-Pro proteases are multifunctional proteins required for potyvirus transmission by aphids (Atreya et ul., 1992; Pirone, 1991) and long-distance spread in plants (Cronin et al., 1995) . By analogy, similar function(s) may he proposed for the B W leader proteinase.

The next downstream ORFs (2, 3, and 4) are arranged as a n overlapping block and encode 6.4-, 65-, and 64-kDa products, respectively (Fig.  2) . The B W 6.4-kDa protein (6.4K) shows marginal similarity to the small hydrophobic proteins encoded in the "triple gene block (TGB) of potex-and carlaviruses (Agranovsky et al., 1991b; Morozov et al., 1989) . However, only a part of the residues conserved in the B W 6.4K and the TGB-encoded proteins may be found in the approximately 6-kDa proteins encoded in CTV, beet yellow stunt virus (BYSV), and L l W (Karasev et al., 199413; Klaassen et al., 1995) . Though their common origin is thus questionable, the small hydrophobic proteins of closteroviruses and potexviruses may be functionally equivalent. The TGB-encoded proteins of potexviruses bind to membranes in vitro (Morozov et al., 1990 ) and mediate the cell-to-cell transport of the viral infection i n vivo (Beck et al., 1991) . The B W 6.4K synthesized in rabbit reticulocyte lysates also showed affinity to cell membranes, and its involvement in the virus infection transport has been suggested (reviewed in Dolja et al., 1994) .

The 65-kDa protein (65K) is strikingly similar to the HSP70 family of cell heat-shock proteins (Agranovsky et al., 1991a) . HSP70s are ubiquitous molecular chaperones which assist proper folding, oligomerization, and transmembrane transport of other proteins (reviewed in Gething and Sambrook, 1992) . Structurally, HSP70s consist of two parts, the N-terminal ATPase domain and the C-terminal peptidebinding domain (reviewed in Craig et al., 1993) . The B W 65K protein contains an N-terminal domain whose sequence and tentative spatial fold are very similar to the HSP70 ATPase, and a unique C-terminal domain that cannot be folded into the P4a (HLA-like) structure typical of the HSP70 peptide-binding domains (Agranovsky et al., 1991a; Rippmann et al., 1991;  F. Rippmann, personal communication) . Hence, the structure of the putative protein-binding domain of 65K suggests a function different from that of classical chaperones. Karasev et al. (1992) first reported that B W 65K expressed in a cell-free transcriptiontranslation system coprecipitates with purified bovine brain microtubules. The binding of 65K was abolished by pretreatment of microtubule preparations with subtilisin, thus suggesting its specificity. Very recently, bacterially expressed 65K and its fragments have been produced in our laboratory (Nikiphorova et al., 1995) . Using a polyclonal antiserum to the C-terminal 13-kDa fragment of 65K, the protein was detected in BYV-infected Tetragonia expansa plants (Agranovsky et al., manuscript in preparation) . I n vitro assays showed that the purified B W 65K, like the cell HSP70s, has magnesium-dependent ATPase activity associated with its N-terminal 40-kDa fragment. However, 65K, unlike its cell homologues, was found to be unable to bind to immobilized denatured protein, and its ATPase activity was not stimulated in vitro by sequence-nonspecific peptides (A. Agranovsky, S. Nikiphorova, 0. Denisenko, and A. Folimonov, unpublished data) . Although these data establish some biochemical characters of the B W 65K pertinent to its function, the possible involvement of 65K in the cell-to-cell movement of the closterovirus infection (Agranovsky et al., 1991a) , which may involve specific interactions with the cell cytoskeleton and translocation machinery , awaits experimental support.

Internal segments in the B W 64-kDa protein and in the equivalent CTV 61-kDa protein reportedly show similarity to a domain in the HSP9O heat-shock proteins Pappu et al., 1994) .

However, the related approximately 60-kDa proteins of LIYV and SPSW fail to display this similarity (Klaassen et al., 1995; author, unpublished observation) .

ORF5 and ORF6 code for the 24K and 22K capsid proteins of BW, respectively (Agranovsky et al., 1991b . The bacterially expressed BYV 24K and 22K share some common epitopes. Upon tissue fractionation, both proteins bulk in the soluble fraction of the BYVinfected cells, but are also found in the cell wall and membrane fractions (Agranovsky et al., 199413) . The structure of the B W virions built of two CPs is reminiscent of some other plant RNA viruses, thus implying functional analogy. First, one cannot but recall the rod-shaped furoviruses and spherical luteoviruses harboring a few copies of CP extended by readthrough of a leaky terminator codon in the CP gene (Bahner et al., 1990; Cheng et al., 1994; Filichkin et al., 1994; Richards and Tamada, 1992) . Notably, such an aberrant protein has recently been mapped to one end of furovirus particles (Haeberle et al., 1994) . The readthrough CP species are held to ensure the persistence of furoviruses and luteoviruses in their respective vectors: fungal zoospores and aphids. Likewise, the semipersistent mode of BYV transmission may be due to the ability of the assembled 24K to cling tightly to cell membranes lining the aphid's alimentary tract. Another (and not necessarily alternative) possibility may be that the 24K tail directs the closterovirus particle to a host (phloem) cell receptor. Conceivably, the fact that p24 is involved in formation of mature virions does not discredit the earlier suggestion that it might participate in the formation of nonvirion ribonucleoproteins adapted for the cell-to-cell transport of genomic RNA (Boyko et al., 1992; Dolja et al., 1994) .

ORFs 7 and 8 encode 20-and 21-kDa products, respectively. The latter is related to a 20-kDa protein encoded in the CTV genome (Pappu et al., 1994) . Apart from this, these products have shown no significant similarities to any proteins in the current database. Recently we produced a polyclonal antiserum against the BYV 21-kDa protein purified from bacteria; using this antiserum to develop Western blots, the 21-kDa protein was detected in soluble and membrane fractions of BYV-infected plants (R. Zinovkin and author, unpublished data).

The German and British strains of B W (BYV-G and BYV-B) have been partially sequenced (Agranovsky et al., 1994a; Brunstedt et al., 1991) , allowing their comparison with BYV-U. BYV-U and BW-G showed 88.5% identity of the nucleotide sequences and the same organization of ORFs 2 to 8 within the 3'-terminal 6-kb region. The majority of nucleotide substitutions in the BYV-G sequence are in the third positions of codons; even when the substitutions change the coding, only about half of the resulting amino acid changes are nonconservative. Nevertheless, the data compiled in Table I1 indicate some differences in the extent of conservation of individual protein sequences in the two B W strains. Proteins 65K, 64K, 24K, and 22K are the best conserved among the strains, whereas the low-molecular-weight proteins (6.4K, 20K, and 21K) are apparently more variable (Table 11 ). There is also a remarkable nucleotide sequence conservation of the intergenic and 3'-untranslated regions among the two strains, suggesting the functional importance of these regions (Table 11 ). The partial sequence of B W -B shows the same disposition of ORFs 4 to 7 (Brunstedt et al., 1991) . The West European strains are apparently closer to each other than to the Ukrainian strain; in particular, the 22-kDa capsid protein sequences of the BYV-G and BYV-B are identical (Table 11 ). 1 -a U, G, and B stand for the Ukrainian, German, and British strains of BW, respectively. The available nucleotide sequences of the BYV-G and BYV-B align with nt 9375-15353 and 11684-14407, respectively, in the complete BW-U sequence.

Percent of amino acid (for polypeptide products) or nucleotide (for noncoding regions) substitutions revealed on pairwise comparisons of strains. Dashes indicate positions where no sequence for the British strain was available. Nontranslated region 1 is between ORFs l b and 2, and nontranslated region I1 is between ORFs 5 and 6.

Among plant RNA viruses, CTV has the largest undivided genome (1 9,296 nt for the Florida T36 isolate), exceeding that of B W by about 4 kb (Karasev et al., 199413, 1995; Pappu ~t al., 1994) . The overall genome structure of CTV is similar to that of BW, comprising the P-Pro, MT, HEL, and POL domains; the small hydrophobic protein; the HSP70 homologue; the 61-kDa protein related to BYV 64K; the 27-kDa CP homologue (27K); the 25-kDa CP; and the 20-kDa protein homologous to the B W 21-kDa protein (Fig. 2 ) . On the other hand, the CTV genome encodes some proteins or polyprotein domains that are not conserved in BW. Interestingly, the P-Pro domain in the ORFla product of CTV is duplicated (Fig. 2; Karasev et al., 1995) . Pairwise comparisons of the putative CTV leader proteins of predicted molecular weights 54 and 55 kDa and the B W 66-kDa protein revealed no sequence similarity among the three proteins apart from the Cterminal 150-residue part encompassing the P-Pro domain. Among the products encoded by the 3'-proximal genes of CTV, the 20-kDa protein (20K) is related t o the B W 21-kDa protein (21K), whereas the 33-, 13-, 18-, and 23-kDa proteins have no homologues in other sequenced closterovirus genomes (Karasev et al., 1995; Pappu et al., 1994) . The 23-kDa protein contains a sequence motif enriched in cystcine and basic residues, which is conserved in putative nucleic acid binding proteins encoded in thc 3'-proximal genes in carlaviruses and allied viruses. Therefore, this putative protein has been implicated in RNA binding and regulation of CTV gene expression .

Severely pathogenic CTV isolates share a common epitope on their particles not found on the particles of mild isolates, thus suggesting the CTV pathogenicity may have some of its determinants associated with the 25-kDa CP (Pappu et al., 1993) . The CTV 27K and 20K proteins were detected in infected plants with polyclonal antibodies against the recombinant proteins (Febres et al., 1994; Pappu et al., 1994) . The bulk of 27K was found in the cell wall fraction of infected citrus leaves, although the protein was also detectable in the soluble and membrane fractions (Febres et al., 1994) . Thus, the 27K association with subcellular fractions differs somewhat from that reported for the homologous B W 24K which, like B W 22K, is predominantly found in the soluble fraction (Agranovsky et al., 1994b) . Recent experiments with yeast two-hybrid system have shown that the CTV 20K (a homologue of the B W 2lK) is capable of homologous interactions, thus suggesting that this protein might function as a di-or multimer (S. Gowda, personal communication).

Sequencing of a 3'-terminal 2.5-kb portion of another CTV isolate, Israeli VT or "seedling yellows" isolate, revealed four ORFs encoding la-, 13-, 20-, and 23.5-kDa proteins that showed close relatedness to the respective products of CTV-T36 (Mawassi et al., 1995a) .

Comparisons of partial sequences of BYSV and carnation necrotic fleck virus (CNFV) indicate their relatedness to CTV and B W Karasev et al., 1994a,b; Klaassen et al., 1995) . Both the BYSV and CNFV genomes contain a conserved array of ORFs coding for POL, a small hydrophobic protein, and a n HSP7O homologue. In addition, the BYSV genome bears a n approximately 30-kDa protein gene inserted between the POL and small hydrophobic protein genes. This configuration is similar to that in the respective part of the CTV genome (Fig. 2) . The fact that B W and CNFV induce very similar patterns of dsRNAs in the infected plant cells (Dodds and Bar-Joseph, 1983 ) is indicative of the overall similarity of their gene layouts.

The genome of lettuce infectious yellows virus (LIYV) is divided among RNA-1 and RNA-2 components of 8.1 and 7.2 kb, respectively (Klaassen et al., 1995) . Interestingly, L I W RNA-1 and RNA-2 show no similarity between their respective terminal untranslated regions (which may be expected to contain putative recognition signals for the replicase), with the exception of the 5'-terminal pentanucleotide, which is identical in both genomic components (Klaassen et al., 1995) . L I W RNA-1 encompasses the overlapping ORFs l a and l b coding for the putative P-Pro and the replicative domains, and the 3'-terminal ORF for a 31-kDa protein (Fig. 2) . Very recently, a full-length cDNA copy of the L I W genome was produced, and it was found that the RNA-1 T7 transcript is necessary and sufficient to support the replication in protoplasts (B. w. Falk, personal communication) . This is the first experimental evidence for the assignment of closterovirus replicative functions to the domains conserved in ORF la/lb. L I W RNA-2 contains genes for the small hydrophobic protein, the HSP7O homologue, the 59-kDa protein distantly related to the B W 64-kDa and CTV 61-kDa products, the 9-kDa protein, the 28-kDa CP, the 52-kDa protein whose C-terminal domain is homologous to the CP, and the 26-kDa protein The monopartite genome of another whitefly-transmissible closterovirus, cucumber chlorotic spot virus (CCSW, has a size of approximately 15.5 kb (Woudt et al., 1993a,b) . The sequence of its coding part shows 5'-terminal overlapping ORFs encoding the domains of the leader P-Pro, MT, HEL, and POL; 3'-proximal ORFs code for the small hydrophobic protein, the HSP70 homologue, the approximately 60-kDa protein, the 9-kDa protein, the 28-kDa CP, the 74-kDa protein containing the C-terminal domain homologous to the CP, and the 23-kDa protein (L. P. Woudt, personal communication). Thus, the undivided genome of CCSV shows an overall arrangement of genes unexpectedly similar to that of the bipartite genomes of L I W and SPSW. In line with this, comparisons of the encoded proteins showed close relatedness among CCSV, LIW, and SPSW, suggesting that the three whitefly-transmissible closteroviruses constitute a distinct evolutionary lineage (Fig. 3) . The LIYV, SPSW, and CCSV ORF l a and l b products (including the N-terminal leader proteins) can be confidently aligned with high statistical scores over almost the entire protein length, whereas their similarity to the equivalent products of B W and CTV is essentially confined to the core replicative domains. The same is true for the encoded HSP70 homologues, the approximately 60-kDa proteins, and the 28K-kDa CPs. The L I W 52-kDa, S P S W 79-kDa, and CCSV 74-kDa capsid protein duplicates show closest relatedness within the approximately 200-residue C-terminal segments, including the CP-like core domains. The gene for a putative 9-kDa protein located upstream of the CP gene is unique for the whitefly-borne closteroviruses; comparison of the 9-kDa sequences encoded in LIW, SPSW, and CCSV showed moderate conservation. The presence of a gene for an approximately 30-kDa product located downstream of the POL gene is common for CTV, BYSV, and the bipartite closteroviruses, although the relatedness of the encoded products is not apparent Karasev et al., 1994b; Klaassen et al., 1995) . The CCSV genome does not encode a product of similar size and location.

The 3l-terminal8.3-kb sequence of the mealybug-transmissible LCV has been recently determined (R. Keim-Konrad and W. Jelkmann, manuscript in preparation). In the 5' to 3' direction, the sequence encompasses the conserved ORFs for the HSP70 homologue, the -60K protein, the 46-kDa (putative) CP, the 76-kDa CP duplicate, and . Trees were constructed using the program PROTPARS of the PHYLIP package (Felsenstein, 1989) from the alignments of 294 and 98 amino acid residues of polymerases and CPs, respectively, excluding the positions containing gaps. The number above each node shows the percentage of bootstrap replicates in which a given node was recovered. Branch lengths are arbitrary. The related sequences of tobacco rattle tobravirus (TRV) polymerase and narcissus mosaic potexvirus (NMV) CP were used as the outgroups in the respective trees. Protein sequences were extracted from database, except for CCSV, SPSW, and LCV (personal communications from L. P. Woudt, U. Hoyer, and W. Jelkmann, respectively). unique ORFs for 21-and 27-kDa proteins (W. Jelkmann, personal communication). The genome of GLRaV-3 reportedly contains the HEL and POL domains, a gene for the HSP7O homologue, and the 43-kDa CP genc (Ling cf nl., 1994) , hut their arrangement has not been yet described. Future sequencing and comparison efforts are expected to elucidate relationships among mealybug-transmissible closteroviruses.

In closterovirus genome organization, variation of a common theme is evident. The invariant elements include the P-Pro, MT, HEL, and I'OL domains; the small hydrophobic protein; the HSP70 homologue, the -60K protein; and the CP and its duplicate. Interestingly, the 65K, 64K and CP genes that are conserved at the species level are apparently least divergent in the B W strains. It is plausible that these genomic elements provide for the characteristic biological patterns common t o all the members of this family. Wherever differences in the conserved genes occur, they intriguingly parallel modifications of' the biological features. For example, the genes for the (major) CP and its duplicate are much more divergent and are transposed in the genomes of whitefly-transmissible CCSV, LIW, and S P S W as compared to aphid-transmissible B W and CTV. In the genome of a mealybugtransmissible LCV, the CP and CP-duplicate ORFs are arranged like those in the whitefly-transmissible species; however, their amino acid sequences are divergent from those of the other family members (Fig. 3) , and the LCV CP is notably large. Could the disparity among the closterovirus CPs and CP duplicates be due to their involvement in and specific adaptation to transmission by different types of vector?

The obvious similarities in gene arrangement and the encoded protein sequences between the whitefly-borne monopartite and bipartite closteroviruses clearly indicate that the lineages represented by CCSV, LIYV, and SPSW on the one hand, and by B W and CTV on the other, had diverged at an evolutionary stage preceding the splitting of the closterovirus genomes. This has important implications in the taxonomy of Closterouiridae; clearly, the second genus (Table I) should include not only the bipartite closteroviruses but also CCSV (and perhaps other whitefly-transmitted closteroviruses). Broadly, sequencing of more closterovirus genomes is expected to shed some light on whether the adaptation to vectors has been a key factor of their molecular evolution.

In closterovirus genomes, the 3'-proximal ORFs vary in number and, as a rule, encode nonconserved protein sequences. The 3'-most ORFs have perhaps diverged most rapidly among the B W strains (Table 11 ). This might indicate that their products are involved in functions connected with fast environmental response, such as modulation of symptom expression or adaptation of the virus to changes in the host or vector populations. A number of B W and CTV isolates have been reported, differing in the severity of symptoms they cause in host plants (reviewed in Moseley and Pappu et al., 1993; Rogov et al., 1993) .

Closterovirus genomes appear to have modular organization (reviewed in Dolja et al., 1994) . In the beginning, the existence of three modules in the B W genome was envisaged, represented by overlapping gene blocks separated by two uridine-rich spacers (Agranovsky et al., 1991b) . It has been suggested that these modules have evolved as distinct entities and that they encode proteins expressed very early (replicase), early (putative transport proteins), and late in the infection (CP and two proteins of unknown function). Revisions of this scheme have been proposed Karasev et al., 1995) ; they agree in placing the genes for the CP and its duplicate into the 3'-terminal module, which seems reasonable. Further, it may be speculated that some closterovirus genes (the approximately 30-kDa ORF and the unique 3'-terminal ORFS) evolved independently of the conserved modules.

Computer-assisted predictions and in vitro experiments have demonstrated that the 295-kDa product encoded in the 5'-most gene of BYV is in fact a polyprotein (Agranovsky et al., 1994a) . The P-Pro domain located in its N-terminal portion mediates autoproteolysis at the Gly-Gly bond to release a 588-residue (66-kDa) leader protein and a C-terminal 229-kDa protein with MT and HEL domains. The catalytic Cys and His residues in the P-Pro active center and its cleavage site (inferred from alignment with potyviral P-Pro domains) have been confirmed experimentally using point mutagenesis and in uitro translation. It has been found that deletion of 245 residues from the N terminus of the leader protein does not impair but rather stimulates the cleavage, and that the His and Cys residues, which are not conserved in the related thiol proteinases, have different effects on autoproteolysis of the B W polyprotein (Agranovsky et al., 1994a) . This is consistent with the results obtained for a related P-Pro of the chestnut blight hypovirulence-associated dsRNA virus (Choi et al., 1991) .

In the CTV ORFla product, the P-Pro domain is duplicated (Karasev et al., 1995) . Based on alignment of the CTV and B W P-Pro sequences, the cleavage sites in the CTV polyprotein have been predicted at the Gly-Gly cioublets at positions 484-486 and 976-977 from the N terminus. Hence, processing of the 349-kDa ORFla product of CTV would yield two leader proteins of 54 and 55 kDa, and a 240-kDa (C-terminal) protein with MT and HEL domains (Karasev et al., 1995) . Putative leader P-Pro domains may also be revealed on computer analysis of the ORFla products of L I W (Klaassen et al., 1995) , SPSW (U. Hoyer, personal communication), and CCSV (L. P. Woudt, personal communication) . The tentative P-Pro cleavage sites in the ORFla products of LIYV, SPSW, and CCSV deviate from the consensus drawn for the BW, CTV, and potyvirus proteinases, being VGIA, LGN, and VGN, respectively. If autocatalysis indeed occurs at these sites, the respective leader proteins should have 412, 496, and 402 residues, respectively.

In the genomes of B W (Agranovsky et al., 1994a) , CTV (Karasev et al., 1995) , CCSV (ten Dam, 1995) , and bipartite closteroviruses (Klaassen et al., 1995;  U. Hoyer, E. Maiss, W. Jelkmann, and J . Vetten, unpublished data), the HEL and POL domains are split between the products of overlapping 5'-proximal ORFs found in 0/+1 configuration, thus indicating that the polymerase may be expressed via +1 ribosomal frameshifting. Although many viral RNA polymerases are expressed as frameshift fusions resulting from translation of overlapping genes, the 0/+1 configuration of the closterovirus replication-associated ORFs is quite unusual [to my knowledge, the only other example is the dsRNA virus of Leishmania (Stuart et al., 1992) l. In all other cases, including retroviruses (Jacks and Varmus, 1985; Jacks et al., 1988) , dsRNAcontaining viruses (Dinman et al., 1991) , and the diverse groups of positive-strand RNA viruses of animals and plants (Brierly et al., 1987; Godeny et al., 1993; Jiang et al., 1993; Makinen et al., 1995; Miller et al., 1988; Xiong and Lommel, 1989) , an upstream ORF and the downstream (POL) ORF are found in 0/-1 configuration. Hence, the tentative frameshifting mechanism in closteroviruses deserves a special comment.

The canonical mechanism of leftward (or -1) frameshifting postulates a one-step-back movement ((Lsimultaneous slippage") of two tRNAs bound to a "shifty" mRNA sequence X XXY YYZ, to decode it a s XXX YYY (Jacks et al., 1988) . The reading-frame switching is stimulated by a pseudoknotted secondary structure (Brault and Miller, 1992; Prufer et al., 1992; ten Dam et al., 1990) ; this effect is probably con-nected with the ability of such a structure to impede the progress of ribosomes along the template (Tu et al., 1992) . Rightward (or +1) frameshifting has been described for the yeast retrotransposons and the E. coli release factor gene (Clare et al., 1988; Craigen et al., 1985; reviewed in Farabaugh, 1993) . In retrotransposons, the frameshifting is enhanced by a rare "hungry" codon adjacent to the shifty codon (Farabaugh et al., 1993) , whereas in the bacterial gene stimulation is provided by the in-frame stop codon and a downstream Shine-Daigarno-like sequence transiently interacting with ribosomal 16s RNA (Weiss et al., 1988) . In short, as follows from comparisons of different frameshifting mechanisms, the reading-frame switching requires some signal(s) to slow down the translating ribosome, thus increasing the chances of the out-of-frame triplet recognition. The B W ORFla ends in a GGGUUUA sequence resembling the "shifty" heptamers of the retroviral type. This resemblance, which we could not but mention in a n earlier work (Agranovsky et al., 1994a) , is probably fortuitous, as such a heptamer is not conserved in the other closterovirus genomes (Fig. 4; Karasev et al., 1995; Klaassen et al., 1995; ten Dam, 1995) . Notwithstanding, we did not suggest the "slippery"consensus to provide for the +1 frameshifting in the B W svstem; rather, our explanation was based on the "U33 grapple" pairing model (Weiss, 1984) . Specifically, offset pairing was postulated between U-7998 in the ORFla UAG stop codon and the nucleotide located leftward to the anticodon of tRNAval to mediate transition of a subset of translating ribosomes into ORFlb (Agranovsky et al., 1994a) . In accord with this, the (G/C)UU U** consensus (where ** designates the last two bases in stop codons) is seen at the 3' termini of ORFla in B W , LIW, SPSW (Fig. 4) , and CCSV (ten Dam, 1995) . Further, it may be speculated that putative secondary structure elements at the B W ORFla stop codon (Agranovsky et al., 1994a) serve to stall the ribosome, thus promoting the frameshifting. At least partially, this RNA fold is conserved in the respective genome regions of CTV and CCSV ( Fig. 5 ; ten Dam et al., 1995) , but not in the LIW genome (Klaassen et al., 1995) .

Alternative frameshifting models have been proposed for CTV and B W (Karasev et al., 1995) and for L I W (Klaassen et al., 1995) . Superposition of the nucleotide and protein sequences in the CTV and BYV HEL/POL gene overlaps reveals a remarkable amino acid conservation profile, suggesting that the frameshifting in the CTV gene occurs after the GUU valine codon, which is not the penultimate triplet there (Fig. 4; Karasev et at., 1995) . By analogy with the yeast retrotransposon system (Farabaugh et al., 1993) , the putative +l frameshifting signal is postulated to be simple, not to include any secondary

guu aAC a a G ucg AgC gau CAC GAc CCG cag CGg GUU-C ucg a U u c G c UCg CAg GCg A W CCU aag AGg AAA ccu gAC u c G ggu A a C uua CAC GAa CCG gcu CGc GUU-C qua g U a a G g U C a CAa GCa AUU CCU cca A G a AAA C m l a / l b Woudt, personal communication) that is similar to those in LIYV and SPSW. structure, and to use codons that may cause ribosome pausing, namely the rare CGG (arginine) codon in the CTV ORFla or UAG stop codon in the B W RNA Karasev et al., 1995) . For LIYV, Klaassen et al. (1995) proposed that the +1 frameshift occurs by slippage of tRNALy" on the AAAG string located eight triplets upstream of the stop codon in the ORFla. Experimental evidence has been obtained for ribosomal frameshifting on expression of closterovirus ORFs l a and lb, and we may not have to wait long for elucidation of the frameshift mechanism. An expression cDNA clone was produced, which contained, in a heterologous context, a 113-nt CCSV-specific fragment encompassing the potential GUU UGA "shifty stop" sequence and a tentative pseudoknot downstream (ten Dam, 1995) . Translation of the SP6 transcript of this clone in wheat germ and rabbit reticulocyte-cell-free systems yielded 35Smethionine-labeled products consistent with the expression of CCSV ORFlb via ribosomal frameshifting with a n efficiency of about 2%. It is worth mentioning that the minimal "shifty stop" sequence of CCSV was found to be incapable of frameshifting in uitro, suggesting that a more elaborate signal must be involved (ten Dam, 1995) . Likewise, we have produced a n expression clone containing the BW-specific insert encompassing the ORFla/lb overlap. Translation of the T7 transcript of this clone in rabbit reticuloctye lysate resulted in ribosomal frameshifting with an efficiency of less than 1% (Agranovsky, Zelenina, and Morozov, unpublished data).

Closterovirus genes located 3'-ward of the POL gene are likely to be expressed via formation of 3'-coterminal subgenomic (sg) RNA species.

Plants infected with CTV, BW, CNFV, SPSVV, LIW, BPW, LCV, and other clostero-like viruses contain a variety of dsRNA species, of which some may correspond to the subgenomic size messengers (Coffin and Coutts, 1992; Dodds and Bar-Joseph, 1983; Gunasinghe and German, 1989; Hu et al., 1990; Larsen et al., 1991; Namba et al., 1991; Winter et al., 1992; K. Eastwell, personal communication) .

In BYV-infected plants, six species of double-and single-stranded RNAs have been identified by Northern blot hybridization, corresponding to the full-sized genomic RNA and to sgRNAs for the 65, 64, 24, 22, and 21K ORFs. The identity of the B W sgRNAs for the 24, 22, and 21K ORFs is supported by in uitro translation of the respective dsRNA species denatured with methyl mercuric hydroxide, which yielded proteins compatible in size with those deduced for the ORF 5, 6, and 8 EVOLUTION OF CLOSTEROVIRUSES 145 products; the product of the most abundant 1.6-kb dsRNA (corresponding to the sgRNA for 22-kDa CP) was immunoprecipitable with a n antiserum to B W particles (Dolja et al., 1990) . None of the B W sgRNAs was found to be encapsidated (Dolja et al., 1990) .

The dsRNA patterns produced in CTV-infected plants vary greatly from strain to strain (Dodds et al., 1987; Guerri et al., 1991; Moreno et al., 1990) . For at least one particular CTV isolate, T36, this pattern was found to be stable when the virus had been propagated in different citrus hosts (Hilf et al., 1995) . Comprehensive Northern blot analysis of single-and double-stranded RNAs from CTV-T36-infected plants demonstrated the presence of nine 3'-coterminal sgRNA species representing the 33, 65, 61, 27, 25, 18, 20, and 23K ORFs (Hilf et al., 1995) . The 3.2-kb (CP) sgRNA of CTV-T36 (Hilf et al., 1995) , as well as the 3.2-, 1.6-, and 0.9-kb sgRNAs coding, respectively, for the CP, 20-kDa, and 23-kDa proteins of CTV-VT (Mawassi et al., 1995a) , were found to be encapsidated. In addition, the CTV-VT encapsidates a 2.4-kb RNA that possesses properties of a defective RNA; as revealed by sequencing, this species is composed of 1.1-kb and 1.3-kb regions derived from the 5' and 3' termini of the CTV genome (Mawassi et al., 199513) . Conceivably, the presence of this defective RNA in virions implies that the putative origin of assembly is in the outskirts of the CTV genome (Mawassi et al., 199513) . The 5' termini of the B W sgRNAs for the major and minor CPs (22 and 24K) were mapped by primer extension to the adenosine residues found 52 and 105 nt upstream of the respective initiating codons (Agranovsky et al., 1994b) . The sequence a t the starts of both sgRNAs of B W is conserved (CCAUUUYA; Y for pyrimidine) and may thus represent a core element of the subgenomic promoter. Interestingly, this element resembles the sequences at the 5' ends of the CP sgRNAs of tobamoviruses and Bromoviridae family members. Bearing in mind that the B W repilcase is most closely related to those of the tobamolike viruses, it is tempting to speculate on parallel conservation, in the process of evolution, of template-binding domains in viral replicases and the signals they recognize in viral RNAs (Agranovsky et al., 1994b) . Recently, we mapped the 5' end of the B W sgRNA for the 64-kDa protein at the adenosine residue located 141 nt upstream of the ORF4 initiating codon (M. Vitushkina and author, unpublished data). The sequence a t the respective start site, ACAUAAUU, significantly deviates from the consensus derived for the 22 and 24K sgRNAs. This, together with the fact that no sequence elements conforming to the CCAUUUYA consensus can be seen in the B W genome sequence upstream of the AUG codons in ORFs 2, 3, 4, 7, and 8, suggests that the 

For genome expression, B W and possibly other closteroviruses combine autoproteolysis by a papain-like proteinase, ribosomal frameshifting, and sgRNA formation, thus resembling the animal viruses belonging to the corona-like superfamily rather than any other known plant virus group (Agranovsky et al., 1994a) . The situation with BYV and CTV, whose genomes contain single and double P-Pro domains, respectively, further parallels that in coronaviruses and arteriviruses, some of which show similar P-Pro duplication (Godeny et at., 1993; Karasev et al., 1995; Lee et al., 1991: Snijder and Horzinek, 1993) . As closteroviruses and corona-like viruses represent evolutionarily disparate lineages (Koonin, 1991; Koonin and Dolja, 1993) , it seems plausible that similar expression strategies in these groups have evolved independently to confer an advantage in expression of large RNA genomes.

Expression of the 5'-proximal genes in closterovirus genomes should produce some proteins in unequal amounts. Thus, translation of the B W genomic RNA should yield the major 295-kDa protein and a fusion 348-kDa protein processed into the 66-, 229-, and 282-kDa proteins. In CTV, translation should result in 349-and 401-kDa polyproteins further processed into the 53-and 54-kDa cleaved leaders, and 240-and 290-kDa proteins (Karasev et al., 1995) . The synthesis of closterovirus ORFla/lb fusion proteins (containing the complete array of replication-associated domains) is perhaps down-regulated (see Section VI,B), as is the case with other virus systems employing translational frameshifting (Brault and Miller, 1992, and references therein) .

In many positive-strand RNA virus genomes, one can discern a trend to regulate the expression of POL and other replication-associated domains (1) by using a leaky nonsense codon or a frameshift signal to isolate the sequence coding for POL from the upstream coding sequence, and (2) by expressing the POL, MT, and HEL (or MT+HEL) domains as distinct products resulting either from polyprotein processing or from translation of individual genomic RNAs. Splitting of the viral replicase into distinct components, whose expression may be regulated separately, is likely to provide the required flexibility in performing different enzymatic functions in RNA replication, namely, unwinding of duplexes, asymmetric synthesis of (+) and (-) strands, synthesis of subgenomic RNAs, and RNA capping (Agranovsky et al., EVOLUTION OF CLOSTEROVIRUSES 147 1994a) . Unlike closteroviruses, all other plant viruses that utilize frameshift for POL expression have small genomes and encode neither MT nor HEL (Koonin and Dolja, 1993) , whereas in corona-like virus genomes both the POL and HEL domains (in this order) are located 3'-ward of the frameshift site (Snijder and Horzinek, 1993) . Thus, closteroviruses are the only viruses known so far in which the frameshift occurs between the sequences coding for HEL and POL.

Comparisons of the MT, HEL, and POL sequences reveal close similarity of closteroviruses to tobamo-, tobra-, furo-, hordei-, idaeo-, bromo-, and ilarviruses, which comprise a compact "tobamo~' lineage (Koonin and Dolja, 1993) within the Sindbis-like supergroup of positive-strand RNA viruses (Goldbach et al., 1991) . Apart from the conserved replicative core that has been vertically inherited from an ancestor shared with tobamo-like viruses, closterovirus genomes show elements of most probably horizontal acquisition (Fig. 6) . This concerns the 65-kDa protein evidently homologous to the HSP7O family of cell chaperones, the CP and its duplicate, and (less likely) the leader P-Pro related to the potyvirus HC-Pro. The capture of foreign genes and intragenomic sequence duplication might be driven by the same mechanism, i.e., copy-choice RNA recombination (Kirkegaard and Baltimore, 1986; Wang and Walker, 1993) . On evolutionary divergence of closteroviruses, some of these elements underwent further shuffling. Thus the sequence coding for the leader P-Pro was duplicated in the CTV genome, and the gene for the CP homologue was extended and moved downstream of the (major) CP gene in the genomes of mono-and bipartite whitefly-borne closteroviruses (or vice versa). The Nand Cterminal domains of the CP homologue of L I W both showed similarity to the CP, so it cannot be ruled out that a triplication of the CP gene has occurred in the L I W genome (Klaassen et al., 1995) . Consistent with this hypothesis, the sizes of the CP duplicates of SPSW (79 kDa) and CCSV (74 kDa), as well as that of LIYV (52 kDa), are rough multiples of their CP sizes (-28 kDa), and the repeated segments of marginal similarity to the CP core may also be found in the N-terminal parts of the 79-and 74-kDa proteins (author, unpublished observation). Thus, expansion of the closterovirus genomes may be partly attributed to insertions and tandem duplications at both ends of the conserved replicative core. Fig. 2 . Asterisk marks the leaky termination codon in the TMV replicase gene; 30K, transport or movement protein gene of TMV; CP, capsid protein.

Another large insertion in the closterovirus genomes lies within the replicative core, between the MT and HEL domains (Fig. 6) . The sequence of this region is significantly diverged in the BW, CTV, and L I W replicases (Dolja el al., 1994; Karasev et al., 1995; Klaassen et al., 1995) . Interestingly, among closteroviruses and related plant viruses the size of the MT-HEL span grows almost linearly with the increase of the genome size; in the case of viruses with divided genomes, it is related rather to the size of the largest genomic component (Table 111) . Generally, a rule "the larger the genome, the larger the replicase" inferred from these comparisons may also be applicable to arteri-, toro-, and coronaviruses of animals (den Boon et al., 1991; Godeny et al., 1993; Snijder and Horzinek, 1993) . This relationship is not trivial, as the overall increase in genome size in these cases is not solely due to insertions in the replicase gene(s), but also to the appearance of new coding sequences flanking the replicative core. It seems quite likely that, in the process of evolution, expansion of closterovirus genomes was attended by an increase in the size of their replicases (Agranovsky, et al., 1994a) . At least one apparent obstacle in maintaining large RNA genomes must be accumulation of mutations in the progeny strands due to low fidelity of viral RNA polymerases (Holland et al., 1982; Steinhauer and Holland, 1987) . It would be interesting to see if, for example, the inserted domains or the leader proteins serve as "spell- a The MT-HEL distance is measured between the C-terminal part of methyltransferase motif IV (Rozanov et al., 1992) and the GKSlT signature in helicase motif I (Gorbalenya and Koonin, 1993) . The HEGPOL distance is measured between the GKSlT and the GDD polymerase signature (Kamer and Argos, 1984) .

For BMV and BSMV, in which MT-HEL and POL domains are assigned to two individual proteins, sizes are given for both the putative methyltransferase-helicase and polymerase.

The sizes of closterovirus replicases are given assuming translational frameshifting for ORFs l a l l b expression and the cleavage of the N-terminal leader protein.

\L checkers" on strand copying. In DNA-dependent DNA polymerases, the 3'-5' exonuclease activity assigned to a distinct protein domain is crucial for high replication fidelity (reviewed in Kunkel, 1988) . Also, these domains might mediate homologous recombination between the virus RNA molecules to get rid of incorrigible errors, thus maintaining viable progeny. In coronavirus replication, recombination is believed to be a key mechanism to combat high-frequency errors (Jarvis and Kirkegaard, 1991; Lai, 1990) . Finally, there is a possibility that the insert between MT and HEL contains a set of distinct domains to recognize the replication signals on a n RNA template. The fact that up to six and nine sgRNA species may be synthesized on BYV and CTV infection, respectively, as compared with only two sgRNAs in the case of TMV, may again be a corollary to the increased complexity of the closterovirus replicases. Naturally, the possibilities mentioned above do not exclude one another.

I n sharp contrast to the MI-HEL span, the distance between HEI, and POL is essentially the same, about 650 residues, in all the tobamolike virus replicases, despite the POL expression mode (Tahle 111). Conservation of this arrangement may reflect constraints imposed on the replicase architectonics that must ensure concerted action of the strand-separating helicase "wedge" and the copying polymerase unit. The fact that the HEL and POL domains are found in two distinct gene products of bromo-and hordeiviruses does not contradict this rule a s it would seem; a t least for brome mosaic bromovirus, it has been demonstrated that the helicase-like and polymerase-like proteins form a complex in which the HEL and POL domains are juxtaposed in a fashion very similar to that in the TMV replicase .

Apart from replication constraints, evolution of closteroviruses toward increasing the genome size would have had to overcome packaging constraints. Comparisons of particle and genome structure of closteroviruses with those of other plant RNA viruses reveal some tendencies that may help one imagine how this could happen. Mono-and multipartite RNA viruses can be subdivided into those having "compressed" and "stretched" genomes, and this may be related to the virion type. Some spherical viruses, namely luteoviruses, tombusviruses, and tymoviruses, have compressed monopartite genomes in which ORFs extensively overlap to form "douhledecker" gene arrangements (Miller ct al., 1988; Morch et al., 1988; Rochon and Tremaine, 1989) . Such economical use of the coding sequence perhaps reflects a compromise between the necessity to widen the repertoire of viral genes and the limited size of' a n RNA molecule that would fit a spherical particle (Bransoin et al., 1995) . However, the maintenance of overlapping genes has a n apparent drawback, as this precludes each gene from being optimally adapted (Keese and Gibbs, 1992) . The genome splitting seen in many RNA virus groups perhaps allows lifting of packaging constraints and minimizing the use of overlapping ORFs (or decompressing the preexisting gene overlaps). Thus, in spherical comoviruses, nepoviruses, dianthoviruses, and Bromoviridae members, the genomes are divided among separately encapsidated RNA components, each containing nonoverlapping gene(s) (Fig. 7) .

An elongated helical capsid is less restrictive for the size of enveloped RNA. As proposed for the corona-like viruses, transition from a spherical (arterivirus-type) to helical (coronavirus-type) nucleocapsid exempted a progenitor of the toro-and coronaviruses from packaging constraints, thus allowing a nearly twofold genome expansion (Godeny et al., 1993) . Characteristically, most of the plant RNA viruses with elongated particles (tobamo-, tobra-, furo-, poty-, carla-, and closteroviruses) have stretched genomes, with modest overlapping of genes if any (Fig. 7) . However, a subset of these viruses having rigid rodlike particles do not encapsidate RNA molecules of more than approximately 7 kb, possibly because of restrictions imposed on the particle length by sterical hindrances in the cell and/or their mechanical fragility. This could have forced the splitting of the genomes of tobra-, hordei-, and furoviruses. In contrast to tobamoviruses having a monopartite 6.4-kb genome with only three genes, these viruses have genomes of 9-10 kb that encompass four to seven genes (Fig. 7) . It has been reasonably hypothesized that acqdisition of flexible and superflexible helical capsids by the ancestors of carlaviruses, potyviruses, and closteroviruses allowed their genomes to grow to 10 kb and 20 kb, respectively . Again, these viruses possess stretched monopartite genomes (Fig. 7) . As for the possible relationship between genome division and capsid type, the existence of bipartite filamentous viruses allied with the last two groups (bymoviruses and bicomponent closteroviruses) suggests the involvement of evolutionary factors other than packaging constraints that might have driven the genome splitting. In closteroviruses, the capsid evolution was crowned by employing the second CP. This conferred on their particles a structural complexity unprecedented among simple elongated viruses, which may be expected to require unusual assembly mechanisms.

The borrowing from Boris Pasternaks book of verse ("Over the Barriers," 19 14-1916) in the title emphasizes that closteroviruses evolved by surmounting the restraints imposed on the genome and particle structure of positive-strand RNA viruses. Closteroviruses have large RNA genomes whose size and coding potential may only be compared to those of the corona-like viruses. Nevertheless, despite similar expression strategies and genome layouts developed in these two groups, closteroviruses cannot be considered as "plant coronaviruses" of a kind, since these similarities do not extend to amino acid sequences. Rather, they reflect independent adaptation to handling large RNA genomes in the two evolutionarily distant lineages. Colinearity and conservation of the main replicative domains clearly suggest the common ancestry of closteroviruses and other plant tobamo-like viruses. However, closteroviruses have followed a distinct evolutionary pathway that has led to dramatic expansion of their genomes. Along with this, their evolution would have had to solve problems connected with replication and packaging of large RNA molecules; it is plausibe that this has been achieved by increasing the size (and functional complexity) of RNA replicase and by using a superflexible capsid made up of two CPs. Expansion of the closterovirus genomes has partially resulted from RNA recombination. It is possible that the horizontally acquired elements brought in novel enzymatic activities and structural elements advantageous for closterovirus adaptation to a distinct ecological niche, distinguished by the phloem-limited nature of infection and the semipersistent mode of insect transmission. In this respect, the most intriguing products are the HSP7O-related protein, having the properties of a microtubulebinding ATPase, and the capsid protein duplicate involved in particle formation. There are many more closterovirus gene products whose functions remain enigmatic, since they have neither sequence-related counterparts in a current database nor known functional motifs. Hence, we may have more surprises. At present, we are making only the first steps in perceiving how closterovirus infection proceeds at the molecular level, despite some progress that has made it possible at least to address these questions. Further studies of the functions encoded by the large RNA genomes of closteroviruses are expected to provide a better understanding of the molecular mechanisms of their interactions with the genomes of their hosts and vectors.

Proc

Mrology 191, Bahner, I

CMI/AAB Descript. Plant Viruses

CMIiAAB Descript. Plant Viruses

Proc

Proc. Natl. Acad. Scz

Plant Viruses

New Aspects of Positive-Strand RNA Viruses

The Plant Viruses. The Filamentous Plant Farabaugh

PHYLIP 3.2 Manual

CMUAAB Uescript. Plant Viruscs. No. 136. Inouye, T

Proc. Nutl

The Plant Viruses. The Filamentous Plant Viruses

Handbook of Plant Virus Infections and Comparative Diagnosis

The Plant Viruses. The Filamentous Plant Viruses

The Plant Viruses. The Filamentous Plant Viruses

Dokladi Rossiiskoi Akademii Nauk

I n "The Togaviridae and Flaviviridae

Proc. Natl. Acad. Sci. ten Dam

Zn "The Plant Viruses. The Filamentous Plant Tollin

Proc. Int. Congr Woudt

Plant Pathol., 6th. Montreal

Yes Yes References Hill and Zetler (1973) Liu and Duffus (1990) Bar-Joseph and Hull (1974) . Duffus (1972) Nakano and Inoye (