Nonstructural proteins of Tobacco rattle virus which have a role in nematode-transmission: expression pattern and interaction with viral coat protein

Tobraviruses are transmitted naturally by root-feeding, ectoparasitic plant nematodes of the genera Trichodorus and Paratrichodorus in a noncirculative manner. The three viruses that constitute the genus Tobravirus, Tobacco rattle virus (TRV), Pea early browning virus (PEBV) and Pepper ringspot virus (PepRSV), have a bipartite, single-stranded RNA genome of positive polarity which is encapsidated into rod-shaped particles. RNA 1, which is highly similar between isolates from the same subgroup, encodes the replicase and movement proteins and a small nonstructural protein with a putative function in seed transmission. In contrast, RNA 2 is highly variable in size and nucleotide sequence among different isolates, even within the same subgroup. In addition to the coat protein (CP), RNA 2 may encode one to three nonstructural proteins (Harrison & Robinson, 1986 ; Visser et al., 1999a ).

Nematode transmission studies with pseudorecombinant TRV isolates produced from the nontransmissible isolate PLB and the transmissible isolate PpK20 showed that vector transmissibility segregated with RNA 2 (Ploeg et al., 1993 ). A mutational analysis of genes in RNA 2 of isolates TRV-PpK20 and PEBV-TpA56 showed that CP and nonstructural proteins are both involved in the transmission process. In addition to the 5' proximal CP gene (i.e. the 2a gene), TRV-PpK20 RNA 2 contains the 2b and 2c genes encoding a 40 kDa protein (40K) (formerly designated as 29·4K) and a 32·8 kDa protein (32·8K), respectively (Hernández et al., 1995 ; Visser et al., 1999b ). Deletions in the 40K gene interfered with transmission of TRV-PpK20 by its vector nematode Paratrichodorus pachydermus, whereas deletions in the 32·8K gene did not (Hernández et al., 1997 ). It was suggested that the 32·8K gene could be involved in transmission of TRV-PpK20 by other vector nematode species. In RNA 2 of PEBV-TpA56, the CP gene is followed by reading frames encoding 9K, 29K and 23K proteins and mutation of each of these reading frames affected transmission of the virus by the vector nematode Trichodorus primitivus (MacFarlane et al., 1996 ; Schmitt et al., 1998 ).

An intriguing aspect of the transmission process is the high specificity in virusvector associations. This concept of specificity is defined with the terms exclusivity, in which there is an apparent unique association between a virus isolate and a vector species, and complementarity, in which different virus isolates may share the same vector species, or conversely, one particular virus isolate may be transmitted by several nematode species (Vassilakos et al., 1997 ; Brown et al., 1995 ; Brown & Weischer, 1998 ). A correlation was found between the serotype of a particular Tobravirus isolate and its vector nematode specificity, suggesting that features on the virus particle determine specific recognition by the vector (Ploeg et al., 1992 ). Studies on the particle structure of tobraviruses by nuclear magnetic resonance spectroscopy revealed that a protruding, mobile segment was located at the C terminus of the CP subunit (Mayo et al., 1993 ). Partial deletion of this mobile segment interfered with transmission of PEBV by its vector nematode (MacFarlane et al., 1996 ).

The requirement for helper protein(s) in vector transmission is not uncommon for plant viruses that are transmitted in a noncirculative manner. The role and function in aphid transmission of HC-Pro of potyviruses and ATF of caulimoviruses has been thoroughly studied (for reviews see Hull, 1994 ; Schmidt et al., 1994 ; Gray, 1996 ; Pirone & Blanc, 1996 ). Present knowledge about the function of these helper proteins of poty- and caulimoviruses supports the bridge hypothesis initially proposed by Govier & Kassanis (1974) . The helper proteins would contain two distinct functional domains, one interacting with the virus particle and the other with a specific receptor site in the aphid stylet or foregut. As suggested by Hernández et al. (1997) , similarities between the mode of tobravirus transmission by nematodes with insect vectors of poty- and caulimoviruses may reflect a common molecular mechanism of virus transmission. The 40K protein of TRV-PpK20 and the RNA 2-encoded nonstructural proteins of PEBV-TpA56 could act as helper proteins with functions in nematode transmission similar to that of HC-Pro and ATF in aphid transmission of poty- and caulimoviruses.. So far, possible interactions between CP and putative tobravirus helper components, or between helper components and nematodes, have not been investigated.

Here, we showed that the TRV-PpK20 40K protein is expressed in leaves and roots of infected plants and that this expression is correlated with the expression of the CP gene. By using the yeast two-hybrid system, we observed that CP of TRV-PpK20 interacts with the 40K and 32·8K proteins and that the C terminus of CP is specifically involved in these interactions.

Plasmids.
The 2a (40K) and 2b (32·8K) genes were amplified by PCR using the primer combinations Bio 186Bio 384 and Bio 187Bio 104 (Table 1), respectively, and with pCaK20-2T7 as a template (Hernández et al., 1996 ). PCR products were subcloned into plasmid pUC21 to allow further cloning of the TRV proteins as C-terminal, in-frame fusions to glutathione S-transferase (pGEX-2T expression system, Pharmacia) by using BglII and EcoRI restriction sites from the pUC21 polylinker. The same subclones were used to clone the 2a and 2b genes as C-terminal, in-frame fusions to the GAL4 DNA-binding domain (B) and transcriptional activation domain (A) of, respectively, the yeast shuttle vectors pAS2-1 (Clontech) and pACT-II (Durfee et al., 1993 ), using an NcoI site that had been introduced by PCR at the initiation codons of the 2a and 2b genes, and BamHI from the pUC21 polylinker. The resulting plasmids were named pA40 and pA32.8 (pACT-II clones) and pB40 and pB32.8 (pAS2-1 clones). A PCR fragment containing the CP gene with an NcoI site at the start codon was generated using primers Bio 185 and Bio 102 (Table 1) and inserted into pAS2-1 and pACT-II using NcoI and BglII (position 1197 of TRV-PpK20 RNA 2), yielding plasmids pB-CP and pA-CP. Plasmids pA-CPΔ19 and pACPΔ79 were obtained in a similar way, using CP fragments cut with NcoI and PvuII (position 1112 of TRV-PpK20 RNA 2) or NcoI and ScaI (position 931 of TRV-PpK20 RNA 2), and introduced in NcoI/SmaI-digested pACT-II. The various CP constructs are schematically shown in Fig. 5. All constructs used in this study have been sequenced to check for unwanted nucleotide mutations caused by the PCR reactions.

Table 1. Nucleotide composition of primers used in this study

(13K): Fig. 5. Mutant CP genes fused to the GAL4 activation domain in the yeast shuttle vector pAS2-1. Vector sequences are represented by dotted lines and viral sequences by continuous lines. Restriction sites used in cloning of viral sequences are indicated. pA-CP, full-length CP gene from AUG start codon to UAG termination codon; pA-CPΔ19, C-terminal 19 aa replaced by 15 nonviral aa; pA-CPΔ79, C-terminal 79 aa replaced by 10 nonviral aa; pA-(empty), empty vector.

Expression of GST fusion proteins in E. coli.
Purification of GST fusion proteins was done according to Frangioni & Neel (1993) . A 100 ml culture of E. coli, containing plasmid pGEX-2T with the 40K or 32·8K gene insert (pGEX32.8 or pGEX40, respectively) was grown to mid-exponential phase and protein expression was induced for 3 h with 0·1 mM IPTG at 37 °C. Cells were pelleted by centrifugation, washed once in ice-cold STE buffer (10 mM Tris, pH 8·0, 150 mM NaCl, 1 mM EDTA) and resuspended in 2 ml STE buffer containing 100 µg/ml lysozyme. Dithiothreitol (DTT) was then added to a final concentration of 5 mM. Lysis was achieved by sonication on ice for approximately 3 min. GST fusion proteins appeared to be present in inclusion bodies that were pelleted by centrifugation at 10000 g and resuspended in 2 ml STE buffer containing 5 mM DTT. Inclusion bodies were solubilized by addition of N-laurylsarcosine (Sarkosyl) to a final concentration of 0·5%. Insoluble debris were removed by an additional centrifugation step. Proteolytic cleavage of the GST fusion proteins to release the 40K and 32·8K proteins from the GST fusions was achieved with thrombin in a phosphate buffer (pH 7·4) containing 0·1 mM CaCl₂ and 150 mM NaCl. The resulting protein extracts were run on 11% SDSpolyacrylamide gels and stained with Coomassie brilliant blue (Sigma).

Preparation of antisera.
Bands containing 32·8K or 40K proteins were excised from Coomassie-stained polyacrylamide gels, homogenized in Freunds incomplete adjuvant and used for the immunization of rabbits (Eurogentec). Injections were subcutaneous, the first with 100 µg of purified proteins (day 0), followed by three boosts, each with 50 µg of proteins at intervals of 14, 14 and 18 days. Blood samples were taken at days 0 (pre-immune serum), 38, 66 and 80 (final bleeding). Antibody titres of all samples were determined by ELISA (Sambrook et al., 1989). For use in Western blot analysis, sera with the highest titres were partially purified by cross-absorption with acetone-precipitated proteins from healthy plants, as described by Hay et al. (1994) .

Immunodetection of viral proteins in plants.
Total protein extracts from roots or leaves of Nicotiana benthamiana plants infected with TRV-PpK20 were prepared as described by Hernández et al. (1996) . To obtain subcellular protein fractions, 1·0 g of leaf or root material was collected 6 days after inoculation and homogenized with a mortar and pestle at 4 °C in 3·0 ml of homogenization buffer (100 mM TrisHCl, pH 8·0, 10 mM KCl, 5 mM MgCl_2, 10% glycerol, 10 mM β-mercaptoethanol) and centrifuged at 1000 g for 15 min. The supernatant was centrifuged at 30000 g for 30 min at 4 °C (Angenent et al., 1989 ). Proteins in the 1000 g pellet, 30000 g pellet and 30000 g supernatant were solubilized in Laemmli buffer. For the detection of CP and 40K protein by Western blot analysis (Towbin et al., 1979 ), gels were loaded with amounts of the subcellular fractions corresponding to 0·5 and 2·0 mg of fresh leaf material, respectively. The blots were analysed with partially purified antisera against CP of TRV-PLB (Angenent et al., 1989 ) and 40K protein of TRV-PpK20 (this work).

Yeast two-hybrid experiments.
Yeast strain pJ69-4A (containing the HIS3 and ADE2 reporter genes; James et al., 1996 ) was transformed with a pACT-II-derived construct, a pAS2-1-derived construct, or with a mixture of a pACT-II- and a pAS2-1-derived construct, as described by Gietz et al. (1995) . pACT-II and pAS2-1 contain Leu and Trp selection markers, respectively. Single transformants were identified by plating on 2% sucrose-containing SC media, lacking leucine (SC-Leu) for pACT-II-derived constructs or lacking tryptophan (SC-Trp) for pAS2-1-derived constructs. Double transformants were selected on SC-Leu-Trp double selective medium.

To identify proteinprotein interactions resulting in the expression of the HIS3 and ADE2 reporter genes (Fields & Sternglanz, 1994 ), four yeast colonies of the transformants listed in Table 2 were plated on media lacking leucine, tryptophan, histidine and adenine (SC-Leu-Trp-His-Ade) and evaluated for their ability to grow at 30 °C. Colonies of double transformants were considered positive when at least three out of four colonies showed significant growth within 5 days after plating on SC-Leu-Trp-His-Ade medium. Positive transformants were plated on SC-Leu-Trp-His-Ade medium containing 10 mM 3-amino-1.2.4-triazole (3-AT, Sigma) to suppress leaky expression of the HIS3 gene (Durfee et al., 1993 ).

Table 2. Interactions between CP and the 32·8K and 40K nonstructural proteins of TRV isolate PpK20

Production of antibodies against the 40K protein
GST fusions of the 40K and 32·8K proteins encoded by TRV-PpK20 RNA 2 accumulated in E. coli as insoluble inclusion bodies that could be purified by differential centrifugation. Fig. 1(a) shows a stained gel run with the GST40K protein (lane 2) and GST32·8K protein (lane 4) after solubilization of the inclusion bodies in buffer containing Sarkosyl. Digestion of the fusion proteins with thrombin resulted in cleavage into GST and 40K or 32·8K proteins (Fig. 1a, lanes 3 and 5). The 40K and 32·8K proteins were excised from the gels and approximately 500 µg of each protein was used to immunize rabbits (two rabbits per protein). The antisera raised against the 40K protein detected the purified GST40K protein in ELISA when diluted 1:20000 (not shown), and specifically detected the 40K and GST40K proteins in Western blot analyses (Fig. 1b, lanes 2 and 3). However, sera from the two rabbits immunized with the 32·8K protein did not detect the 32·8K or GST32·8K protein in ELISA or in Western blot analyses (results not shown). Only the antiserum against the 40K protein was used in further studies.

(58K): Fig. 1. Expression of 32.K and 40K proteins in E. coli. (a) Coomassie blue-stained gel run with partially purified E. coli extracts containing GST40K and GST32·8K fusion proteins before (lanes 2 and 4, respectively) and after (lanes 3 and 5, respectively) digestion with thrombin. (b) Western blot analysis of E. coli-expressed GST40K protein before (lane 3) and after (lane 2) digestion with thrombin. The soluble fraction (lane 4) and 10000 g pellet fraction (lane 5) from untransformed E. coli cells were used as controls. The blot was incubated with the antiserum against the 40K protein in a 1:2000 dilution. Lanes 1 of panels (a) and (b) were loaded with marker proteins; molecular masses are indicated on the left.

Expression of CP and 40K protein in plants
The accumulation of the 40K protein in virus-infected N. benthamiana plants was studied in a time-course experiment. After inoculation, samples of systemically infected leaves from four plants were taken daily and pooled. Total protein extracts were prepared and analysed by Western blotting using antisera against CP (Fig. 2a) and 40K protein (Fig. 2b). Accumulation of both proteins became detectable 4 days post-inoculation (p.i.) and reached a maximum at 67 days p.i. At 8 and 9 days p.i. the level of the 40K protein declined more rapidly than the level of CP. In contrast to CP isolated from purified virus particles, CP from total plant extracts migrated in the gel as a double band. This phenomenon has been observed previously (Angenent et al., 1989 ).

(28K): Fig. 2. Accumulation of CP and 40K protein in systemically infected leaves of N. benthamiana plants. Leaf samples were analysed by Western blotting at days 19 (lanes 19) after inoculation of the plants with TRV-PpK20, using antisera against CP (a) or 40K protein (b). Lanes 0, healthy plants. The positions of CP and 40K protein are indicated on the right. Fig. 3. Accumulation of CP and 40K protein in the roots of infected N. benthamiana plants. Six days after inoculation of the leaves of plants with TRV-PpK20, root samples from six plants (lanes 16) were analysed by Western blotting, using antisera against CP (a) or 40K protein (b). Lanes 7 were loaded with CP purified from virus particles and 40K protein partially purified from E. coli. The positions of CP and 40K protein are indicated on the right.

Involvement of the 40K protein in nematode transmission of the virus would require its accumulation in the root system of the plants. Six N. benthamiana plants were inoculated with TRV-PpK20 and after 6 days protein was extracted from the root system of the plants and analysed by Western blotting for the presence of CP (Fig. 3a) and 40K protein (Fig. 3b). CP and 40K protein were readily detectable in the roots of all six plants but the relative accumulation of the two proteins was variable. For instance, roots of the plant analysed in lane 4 of Fig. 3 contained a relatively high level of CP and a relatively low level of 40K protein; the reverse is observed in the roots of the plant analysed in lane 5, whereas relatively high levels of both proteins were present in the roots of the plant analysed in lane 3.

Subcellular localization of CP and 40K protein
To investigate the subcellular localization of CP and 40K protein, virus-infected and non-infected leaf and root samples were homogenized and the homogenate was fractionated into a 1000 g pellet, a 30000 g pellet and a 30000 g supernatant. The 1000 g pellet contains nuclei and chloroplasts, the 30000 g pellet contains membrane fractions whereas cytosolic proteins are present in the 30000 g supernatant (Angenent et al., 1989 ). Fig. 4 shows a Western blot analysis of the presence of 40K protein (Fig. 4a) and CP (Fig. 4 b) in the three subcellular fractions from roots (lanes 2, 3 and 4) and leaves (lanes 5, 6 and 7). The distribution of the 40K protein over the 1000 g pellet, 30000 g pellet and 30000 g supernatant of root and leaf homogenates closely resembled that of CP. This could point to a possible co-localization of 40K protein with (aggregates of) virus particles. Interestingly, the antiserum against the 40K protein specifically detected a 22K protein in the 30000 g supernatant from the roots of infected plants (Fig. 4a, lane 4). As this protein is equally present in the same fraction from the roots of healthy plants (Fig. 4c, lane 4) it is apparently a host protein.

(66K): Fig. 4. Localization of 40K protein and CP in subcellular fractions from roots and leaves of infected N. benthamiana plants. Six days after inoculation of the leaves with TRV-PpK20 (a, b), or after mock inoculation (c), homogenates of the roots (lanes 2, 3 and 4) or systemically infected leaves (lanes 5, 6 and 7) were fractionated into a 1000 g pellet (lanes 2 and 5), 30000 g pellet (lanes 3 and 6) and 30000 g supernatant (lanes 4 and 7), and proteins in these fractions were analysed by Western blotting, using antisera against 40K protein (a, c) or CP (b). Lanes 1 were loaded with marker proteins; molecular masses are indicated on the left. The positions of 40K protein and CP are indicated on the right; the arrow head indicates a 22K plant protein detected by the 40K protein antiserum.

Interaction of CP with 32·8K and 40K proteins
To analyse possible interactions between the proteins encoded by RNA 2 of TRV-PpK20, the CP gene, 40K gene and 32·8K gene were cloned into the yeast shuttle vectors pAS2-1 (containing the GAL4 DNA-binding domain) and pACT-II (containing the GAL4 transcription activation domain) of the two-hybrid system. In addition, the C-terminally truncated CP genes shown in Fig. 5 were also cloned in these yeast vectors. Yeast strain pJ69-4A, which carries the HIS3 and ADE2 reporter genes for interaction, was doubly transformed with the pAS2-1 and pACT-II derivatives in all the combinations listed in Table 2. As a control for possible self-activation of the reporter genes, yeast cells singly transformed with pAS2-1 or pACT-II derivatives containing the CP, 32·8K or 40K genes (data not shown), or yeast cells doubly transformed with one recombinant plasmid and one empty vector were tested for growth on selective medium. None of these controls was found to be self-activating (Table 2). Moreover, no interaction was observed in the two-hybrid system between CP, 32·8K or 40K proteins and the control proteins p53, SV40 T antigen or human Lamin c provided with the Stratagene kit (HybriZap-2.1; data not shown).

No homodimer 32·8K32·8K or 40K40K protein interactions could be observed in the yeast two-hybrid system (Table 2). Also, no heterodimer 32·8K40K protein interaction was detectable, irrespective of whether the proteins were fused to the activation domain or DNA binding domain of GAL4. However, the two-hybrid system did reveal a specific CPCP interaction and interactions between CP and the 32·8K and 40K proteins (Fig. 6 and Table 2). When the selection pressure was increased by addition of 10 mM 3-AT to the medium, yeast colonies with the CPCP interaction were still obtained but no CP32·8K or CP40K protein interactions were observed (not shown). This may indicate that CPCP interactions observed with the two-hybrid system are stronger than CP32·8K or CP40K protein interactions. In the absence of 3-AT, visible growth of yeast colonies with the CPCP or CP32·8K protein combinations was observed after 3 days whereas growth of colonies with the CP40K protein combination was observed after 4 days. This suggests that the CP40K protein interaction is weaker than the CP32·8K protein interaction.

(63K): Fig. 6. Growth of transformed yeast cells after incubation for 6 days at 30 °C on selective medium lacking histidine and adenine. Strain pJ69-4A was doubly transformed with pB and pA derivatives. The pB derivatives contain fusions of the GAL4 DNA binding domain with full-length CP (pB-CP), 32·8K protein (pB-32) or 40K protein (pB-40). The pA derivatives contain fusions of the GAL4 activation domain with the CP constructs shown in Fig. 5.

To identify domains of CP involved in interactions with 32·8K and 40K proteins, two CP mutants with modified C termini were engineered (Fig. 5). Plasmids pA-CPΔ19 and pA-CPΔ79 encode the GAL4 activation domain fused to C-terminally truncated CP. In mutant CPΔ19, the C-terminal 19 amino acids (aa) are replaced by 15 nonviral aa; in mutant CPΔ79, the C-terminal 79 aa are replaced by 10 nonviral aa. Mutant CPΔ19 showed binding to CP and the 32·8K protein but no interaction with the 40K protein was observed in the two-hybrid system (Table 2, Fig. 6). The C-terminal deletion in mutant CPΔ79 did not affect the interaction of the mutant with CP but abolished interactions with the 32·8K and 40K proteins (Table 2, Fig. 6). The failure to detect interactions in the combinations 32·8K/CPΔ79, 40K/CPΔ19 and 40K/CPΔ79 confirms that fusions of CPΔ19 or CPΔ79 to the activation domain of GAL4 did not confer self-activation of the reporter genes. After serial passage of TRV by mechanical inoculation, sequences required for nematode transmission are readily lost from RNA 2 (Hernández et al., 1996 ). Of the several Tobravirus isolates that have been (partially) sequenced, only three are transmissible by vector nematodes: TRV-PpK20, TRV-TpO1 and PEBV-TpA56 (Hernández et al., 1995 ; MacFarlane & Brown, 1995 ; MacFarlane et al., 1999 ). TRV-TpO1 and PEBV-TpA56 are both transmitted by Trichodorus primitivus but not by Paratrichodorus pachydermus, the vector of TRV-PpK20. Currently, it is not known whether vector specificity is determined by CP, nonstructural proteins or both. Sequence identity between CPs of the three transmissible isolates is about 5560% and the isolates belong to different serotypes. The 29K proteins encoded by the 2b genes of TRV-TpO1 and PEBV-TpA56 share about 45% amino acid sequence identity whereas their identity with the 40K protein encoded by the TRV-PpK20 2b gene is less than 20%. No significant similarities were observed between the proteins encoded by the 2c genes of TRV-PpK20 (32·8K), TRV-TpO1 (18K) and PEBV-TpA56 (23K) (MacFarlane et al., 1999 ). Determinants in RNA 2 that are required for transmission of TRV-TpO1 have not yet been analysed. The 2b genes of TRV-PpK20 and PEBV-TpA56 were both shown to be required for transmission of these viruses by their vector nematodes. Although the 2c gene of TRV-PpK20 appeared to be dispensable for transmission of this isolate by P. pachydermus, mutation of the PEBV-TpA56 2c gene strongly reduced the efficiency of transmission of this isolate by T. primitivus (MacFarlane et al., 1996 ; Hernández et al., 1997 ; Schmitt et al., 1998 ). Moreover, RNA 2 of TRV-TpO1 and PEBV-TpA56 contains a reading frame for a 9K protein between the CP and 2b genes that is absent in TRV-PpK20 RNA 2. A frame-shift in the 9K gene, which is in-frame with the CP gene, reduced the efficiency of transmission of PEBV-TpA56 but mutation of its AUG initiation codon did not (MacFarlane et al., 1996 ). However, CP9K fusion proteins have not been detected in Western blots of extracts of PEBV-TpA56-infected plants (quoted in MacFarlane et al., 1999 ). Also, the UAG termination codon of the TRV-PpK20 CP gene is probably not leaky because a mutant with this codon changed into UAA produced the same doublet of CP-specific bands in infected plants as shown in Fig. 3(a) (unpublished results).

Antisera against the PEBV-TpA56 29K and 23K proteins have been used to demonstrate the expression of the 2b and 2c genes of this isolate in leaves and roots of N. benthamiana plants (Schmitt et al., 1998 ). Similar to the results with the PEBV-TpA56 29K protein, we observed that expression of the TRV-PpK20 40K protein in leaves follows that of CP with a slight delay in the start of 40K protein synthesis and a slight advance in the cessation of expression or protein turnover. Another similarity is the observation that CP and 2b gene products of TRV-PpK20 and PEBV-TpA56 accumulated at variable ratios in the roots of infected plants. The reason for this variation is unclear. The distribution of CP over the 1000 g pellet, 30000 g pellet and 30000 g supernatant from extracts of TRV-PpK20-infected N. benthamiana plants (Fig. 4 b) was very similar to that observed previously for corresponding subcellular fractions of homogenized tobacco protoplasts (Angenent et al., 1989 ). In this previous study, the 16K nonstructural protein encoded by TRV-PpK20 RNA 1 was found to be exclusively present in the 30000 g pellet. In contrast, the distribution of the 40K protein over the three subcellular fractions closely resembled that of CP (Fig. 4a). If the 40K protein is physically attached to virus particles, the interaction is apparently disrupted by the virus isolation procedure as no 40K protein was detectable by Western blot analysis in purified virus preparations (results not shown).

In addition to the co-localization of CP and 40K protein in subcellular fractions, an interaction was observed between the two proteins in the yeast two-hybrid system. Although the interaction was weaker than the interaction observed between CP subunits, it was apparently specific. Deletion of the C-terminal 19 aa of CP abolished the interaction with the 40K protein but not the interaction with the full-length CP. From comparisons of tobamoviral CPs, it has been predicted that internal domains mediate interactions between CP subunits in virions (Goulden et al., 1992 ). Replacement of the C-terminal 15 aa of CP of TRV-PpK20 by three nonviral aa did not affect virion formation (Hernández et al., 1996 ). The C terminus of tobraviral CP is predicted to extend away from the virion surface as a flexible arm with a length of 22 aa in TRV-PpK20, 17 aa in TRV-TpO1 and 29 aa in PEBV-TpA56 (Mayo et al., 1993 ; MacFarlane et al., 1999 ). Removal of the C-terminal 15 aa of CP of PEBV-TpA56 (but retaining the terminal alanine residue) abolished nematode transmission of the virus (MacFarlane et al., 1996 ). Our observation that deletion of the C-terminal 19 aa of CP of TRV-PpK20 interfered with its interaction with the 40K protein in the two-hybrid system indicates that a CP40K protein interaction plays a role in the transmission process. This observation is in line with the hypothesis that the 40K protein forms a bridge between putative receptors in the food canal of the vector nematode and specific domains of CP on the surface of virus particles (Brown et al., 1995 ; Hernández et al., 1997 ). However, we cannot rule out the possibility that the nonviral amino acids at the C terminus of CPΔ19 affected the interaction of this protein with the 40K protein. The sequence of the C-terminal 15 aa of CP of PEBV-TpA56 and TRV-TpO1 is almost identical but very different from the C-terminal sequence of CP of TRV-PpK20. It would be interesting to see whether a possible specificity in CP2b protein interactions would correlate with vector specificity of these three viruses.

The 32·8K protein encoded by the TRV-PpK20 2c gene is neither required for transmission by P. pachydermus nor for replication of mechanically inoculated virus, and it has been speculated that this 32·8K protein may play a role in transmission of TRV-PpK20 by vector nematodes other than P. pachydermus (Hernández et al., 1997 ). Because preparation of an antiserum against this protein was not successful, we could not investigate a possible common subcellular localization of CP and 32·8K protein. However, similar to the 40K protein, the 32·8K protein did interact with CP in the two-hybrid system. A CP deletion analysis pointed to a determinant involved in this interaction that is located between aa 19 and 79 from the C terminus.

A direct interaction between the CP and HC-Pro of potyviruses has been demonstrated by using a protein blotting-overlay technique (Blanc et al., 1997 ). In that study, HC-Pro was purified from infected plants as a biologically active dimer (Thornbury et al., 1985 ). By the use of the two-hybrid system, PVA CPCP and HC-ProHC-Pro dimer formation could be demonstrated but no CPHC-Pro interaction was detectable (Guo et al., 1999 ). We did not detect dimer formation of the 40K or 32·8K proteins in the two-hybrid system, indicating that the 40K protein functions as a monomer in transmission of TRV-PpK20 by the vector nematode P. pachydermus.

This work was financially supported by the Association of Biotechnological Research Schools in The Netherlands (ABON).