A discovery 70 years in the making: characterization of the Rose rosette virus

Rose rosette (RR), a disease of unknown aetiology is widespread east of the Rocky Mountains of the USA. Symptoms include excessive lateral shoot growth, excessive thorniness, witches' broom, leaf proliferation and malformation, mosaic, red pigmentation and eventually plant death (Fig. 1; Amrine, 2002). Since the first report in the 1940s (Conners, 1941), there has been little progress towards the identification of the RR agent. Some of the symptoms are similar to those commonly associated with phytoplasmas, but treatment with tetracycline did not cure affected plants, indicative of the non-bacterial nature of the RR agent (Epstein & Hill, 1995). The discovery of double membrane-bound particles (Gergerich & Kim, 1983), the identification of the eriophyid mite Phyllocoptes fructiphilus as a vector (Amrine et al., 1988) and the isolation of dsRNA from infected material (Di et al., 1990) suggested that the disease is caused by a virus.

In this communication, the putative causal agent of the disease, a negative-strand RNA virus provisionally named Rose rosette virus (RRV), is characterized. RRV was detected in 84/84 plants showing typical disease symptoms, collected from the foothills of the Rocky Mountains to the eastern seaboard of the USA. The wide distribution makes the existence of distinct virus isolates with different disease potential quite plausible. This possibility was investigated using two genomic regions of the virus from 22 isolates collected from different states and separated by more than 1300 km.

This study was initiated with the isolation of dsRNA from 12 RR plants by using a modified protocol from Yoshikawa & Converse (1990) (Supplementary Methods, available in JGV Online). After optimization of the dsRNA extraction, a single isolate was selected to serve as a template for virus characterization. Degenerate oligonucleotide-primed reverse transcriptase-PCR (DOP RT-PCR) (Telenius et al., 1992) was used on a cDNA template obtained from dsRNA extracted from symptomatic tissue prepared as described previously (Tzanetakis & Martin, 2008). The sample was then subjected to high throughput sequencing using the Illumina platform. Sequencing was performed at CGRB (Oregon State University) according the manufacturer's recommendations. The contigs were generated by using Velvet (Zerbino & Birney, 2008) and CAP3 (Huang & Madan, 1999), subjected to blastx using blast plan (He et al., 2007). The Illumina runs yielded partial sequences of RNA 1 and 2 of the virus. The missing parts of these molecules were completed by using RT-PCR with specific primers and direct cloning and sequencing (Supplementary Table S1, available in JGV Online).

The remaining genomic molecules of the virus were obtained using, as a guide, the published sequences of European mountain ash ringspot associated virus (EMARaV) (GenBank accession nos NC_013105, NC_013106, NC_013107 and NC_013108) and Fig mosaic virus (FMV; FJ769161, AM941711, FM864225 and FM991954). The first and last 13 nt of EMARaV and FMV share more than 90% identity and this property was used to develop primers that would potentially amplify the complete RNAs of the virus. dsRNA was A-tailed by a using Poly(A) Tailing kit (Ambion) according to the manufacturer's recommendations followed by RT as described previously (Tzanetakis & Martin, 2004). The cDNA was then subjected to PCR by using LA Taq (TakaraBio) and primers designed to match the last 13 bases of the 5′ (5′-AGTAGTGTTCTCC-3′) and 3′ (5′-GGAGTTCACTACT-3′) ends of EMARaV and FMV with an oligoT₁₃ extension (Supplementary Table S1). Amplicons were cloned and sequenced (Supplementary Methods) to obtain at least three times genome coverage.

The first and last 13 bases of the virus RNAs were verified using RACE (Supplementary Table S1). Four RT primers, one primer for each RNA species, were chosen upstream of the PCR primers. Total RNA was used as the template for the 5′ RACE RT (Tzanetakis et al., 2007), followed by poly(C)-tailing. The sample was then subjected to PCR using LA Taq (TakaraBio), the virus-specific primers and a primer for the poly(C)-tail (Supplementary Table S1). For the 3′ RACE A-tailed dsRNA was used as a template for the RT-PCR using an oligo d(T) and a specific primer. All amplicons were cloned and sequenced as before. Programs used for bioinformatics analysis can be found in Supplementary Methods.

RNA 1 is 7026 nt long and contains a single ORF between nt 6919–6917 and 91–89 coding for a protein with a molecular mass of 264.8 kDa (Fig. 2). It encodes a putative RNA-dependent RNA-polymerase (RdRp) with signature RdRp motifs from members of the family Bunyaviridae between residues Lys₁₀₂₆–Asn₁₃₀₄ (Supplementary Fig. S1, available in JGV Online). Motifs A (DXXKWS_1106–1111) and C (SDD_1233–1235) are part of the palm domain of the replicase (Lukashevich et al., 1997; Bruenn, 2003) and are involved in divalent-metal cation binding. Motif B (QGXXXXXSS_1192–1200) is thought to be involved in RNA binding with the conserved Gly₁₁₉₃ residue allowing for mobility in the peptide backbone. The Lys residues in motif D (KK_1285–1286) are thought to have catalytic activity as they are reported to be close to the Asp₁₁₀₆ residue in motif A through the tertiary structure of the protein. Motif E in emaraviruses (EFLST) is similar to the conserved residues in tospoviruses (EFXSE; Bag et al., 2010), located from residues Glu₁₂₉₅ to Thr₁₂₉₈ in RRV; however, there is an amino acid substitution from the polar-neutral Thr to the negatively charged Glu. Motif E is believed to be involved in cap-snatching in bunyaviruses as well as having possible endonuclease activity. The genomic RNA of RRV was found to be uncapped, though the mRNA was capped (data not shown). As a mechanism for cap-snatching is required to accomplish this, it is likely that the substitution from Glu to Thr does not alter the proposed function of motif E. An ATP-dependent helicase motif was located between residues Ile₁₄₁₂ and Glu₁₄₄₅, but its short length suggests that it is non-functional. The RRV RdRp shares 68 % aa identity (83 % similarity) with the FMV and 49 % aa identity (68 % similarity) with the EMARaV orthologues.

RNA 2 is 2245 nt long with a single ORF, starting at nt 2195–2193 and terminating at nt 260–258, encoding a putative glycoprotein precursor, p2, of molecular mass 73.8 kDa. The glycoprotein precursor contains six potential sites for N-glycosylation and is predicted to form homodimers (Mericity; Garian, 2001). Two potential sites for cleavage were predicted using SignalP (Bendtsen et al., 2004) between residues Gly₂₀/Thr₂₁ and Ala₁₉₅/Asp₁₉₆ though the cleavage site at Gly₂₀/Thr₂₁ was supported by a lower score. Cleavage at Ala₁₉₅/Asp₁₉₆ yields two mature glycoproteins, Gp1 of 22.4 kDa and Gp2 of 51.4 kDa. The first 20 aa of Gp1 contain a signal peptide predicted to be secreted to the endoplasmic reticulum. Two transmembrane domains (Sonnhammer et al., 1998) were predicted for Gp1, between residues Val₁₁₄–Leu₁₃₆ and Tyr₁₇₆–Ala₁₉₃; Gp2 contains one transmembrane domain from Asn₃₉₁ to Asp₄₁₃.

The p2 of EMARaV was reported to contain a phlebovirus glycoprotein motif (Mielke & Muehlbach, 2007; Supplementary Fig. S1) and this short motif, GCYXCXXG_481–488, was also found in RRV Gp2. The protein has 51 % aa identity (73 % similarity) with FMV and 40 % aa identity (60 % similarity) with the EMARaV orthologues. Because p2 is predicted to have glycosylation sites and transmembrane domains, as well as a glycoprotein motif and a predicted structure similar to other glycoproteins (PHYRE, Supplementary Methods; Kelley & Sternberg, 2009; Fig. 2), we predict that RRV p2 is a glycoprotein precursor.

RNA 3 is 1544 nt long and contains a single ORF (nt 1445–495) for a putative nucleocapsid protein of 35.6 kDa. The predicted size is similar to that of the orthologues from High plains virus (HPV; 32 kDa), Pigeon pea sterility mosaic virus (32 kDa), EMARaV (35.1 kDa) and FMV (35.1 kDa) p3 orthologues. The putative protein shows identity from 26 % aa (47 % similarity) of the putative nucleocapsid of HPV to 60 % aa (74 % similarity) of FMV p3.

RRV p3 is predicted to form homodimers (Mericity; Garian, 2001) and contains stretches of positively charged residues, predicted to bind RNA (BindN at 85 % specificity; Wang & Brown, 2006). The predicted structure (PHYRE; Kelley & Sternberg, 2009) is bowl-shaped with interlocking protrusions and β-sheets, predicted to bind RNA (Fig. 2). This gives credibility to the prediction that the EMARaV, FMV and RRV p3 orthologues are nucleocapsid proteins. No motifs, identified in other viral proteins were found in p3. However, two conserved regions, (NVVSXXKXXXA_134–143 and NRLA_182–185), reported earlier as possible emaravirus nucleocapsid motifs (Elbeaino et al., 2009) were also found in RRV p3 (Supplementary Fig. S1).

RNA 4 is 1541 nt long and contains a single ORF starting at nt 1458 and terminating at nt 373 (Fig. 2). This ORF encodes a putative 41 kDa protein of undetermined function. The putative protein contains an ATPase motif between Phe₇₈–Asp₁₁₉ and a dnaK motif between residues Asp₃₀₅ and Met₃₅₀ (Supplementary Fig. S1). The protein product shows 59 % aa identity (76 % similarity) to the FMV orthologue. The predicted structure using PHYRE (Kelley & Sternberg, 2009; Fig. 2) shows a horseshoe-shaped monomer that is predicted to form homodimers (Mericity; Garian, 2001). It is possible that p4 also has the peptide binding properties of dnaK-containing heat-shock proteins. Although present in an unrelated group of viruses, it has been proven that closteroviral Hsp70h are involved in virus cell–cell movement (Dolja et al., 2006) requiring both the dnaK and ATPase domains for movement (Alzhanova et al., 2001). This suggests that RRV p4 is involved in cell–cell movement in a manner similar to closteroviruses with p4 threading the nucleocapsid-coated RNAs through the plasmodesmata in an ATP-dependent manner.

The virus RdRp clusters with the RdRps of two members of the newly formed genus Emaravirus (EMARaV and FMV), as supported by absolute bootstrap values (1000/1000 pseudoreplications), providing evidence that RRV is a new member of the genus (Fig. 3).

Koch postulates could not be fulfilled for the virus as there were no local lesion alternative hosts identified and virus purifications were unsuccessful (Laney, 2010). For this reason, symptomatic tissue from 84 cultivated and R. multiflora roses was collected from nine states. Thirty asymptomatic roses were used in this study as negative controls. After total nucleic acid extraction, cDNA was synthesized and subjected to PCR using detection primers developed by using Primer-blast (Rozen & Skaletsky, 2000) (Supplementary Methods). RRV was detected in all symptomatic roses from all nine states, but not in any of the asymptomatic plants (Supplementary Table S2, available in JGV Online). Given the strong correlation between virus presence and disease and the absence of the virus in asymptomatic plants collected from areas of extreme disease incidence, it is most probable that RRV is the causal agent of the disease.

The diversity of RRV p3 and p4 was investigated using 22 isolates from eight different states. The coding regions were obtained using PCR, with the product being directly sequenced. All respective sequences were aligned with clustal w (Thompson et al., 1994). The nucleotide identities for both regions ranged from 97 to 99 %. Most of the isolates shared 97 to 100 % aa identity for both regions with two exceptions. For p3, one isolate from Missouri, contained an extra Gly at position 170. The amino acid identity for this isolate compared with the others ranged from 93 to 94 %. Additionally, one isolate from Mississippi, had an amino acid deletion at the C terminus of p4, conserved as Asn₃₄₉ in the other isolates. The amino acid identity, compared with the other isolates, for this isolate ranged from 94 to 96 % (Supplementary Tables S3 and S4, available in JGV Online).

Geographical location did not influence diversity between RRV isolates. Isolates that were geographically close could be more diverse than isolates found hundreds of kilometers apart. There are three possible explanations for this observation: it may be that RRV only infects Rosa spp. minimizing any host driven diversity, unlike situations of viruses with a wide host range where different isolates/strains evolve to adapt to different hosts. A survey of 34 plant species located around symptomatic roses in different states yielded no positive results for an alternative host (Laney, 2010). It may also be that the virus replicates in the mite vector, which could provide an evolutionary bottleneck as only variants that can replicate in both plant and mite are transmissible. Another possible explanation is that the low diversity is reflective of emaraviruses. Similar results were observed when a diversity survey was performed for EMARaV in Europe. Sequencing the coding region for RNA 3 showed that isolates shared identity between 97 and 99 % (Kallinen et al., 2009). Again in this case geographical location did not influence diversity; isolates from Germany and Finland were more conserved than isolates from adjacent trees in Finland.

This communication gives an answer to a question lingering for more than 70 years: what causes RR? The absolute association of a new emaravirus found in 84/84 symptomatic plants, but not in any of the 30 asymptomatic plants, collected from areas more than 1300 km apart, indicates that RRV is most likely the causal agent of the disease. The discovery of the novel virus, provisionally named RRV, and the development of a reliable detection protocol is of great importance to the nursery and the ornamental industries around the world as it is now possible to test for the virus and freely move material within and between countries without the fear of introducing the disease into areas where it is not present.