Microsoft Word - 2012-Khalifa-Plant Pathol

The pathogenicity properties of Potato virus Y (PVY; genus Potyvirus, family Potyviridae) isolates collected in naturally-infected pepper (Capsicum annuum) fields in Tunisia were evaluated against recessive resistance alleles at the pvr2 locus of pepper. Two pathotypes were observed. Pathotype (0,1,3) isolates were able to infect plants carrying the susceptibility allele pvr2, together with pvr2/pvr2 and pvr2/pvr2 plants but not pvr2/pvr2 plants. Pathotype (0) isolates were only able to infect pvr2/pvr2plants. On the other hand, sequence data and phylogenetic analyses revealed three major groups of isolates, each characterized by particular amino acid residues in the central part of the VPg, the pathogenicity factor towards pvr2. Correspondence between pathogenicity properties and phylogeny suggested a single evolution step for pathogenicity towards the pvr2 and pvr2 resistances, possibly under the selective pressure of pvr2. Indeed, 23% of the pepper plants in this area were shown to carry the pvr2 resistance while pvr2 was not detected. The data suggested that pathogenicity towards pvr2 and pvr2 were not costly for PVY to infect susceptible pepper genotypes and supported the matching allele model for pepper-PVY interactions. INTRODUCTION Potato virus Y (PVY; genus Potyvirus; family Potyviridae) is an important pathogen in solanaceous crops including potato, pepper, tomato and tobacco. PVY isolates have been classified into three main phylogenetic groups, named groups O, N and C, which possess particular host range properties. Almost all pepper isolates of PVY belong to the C group, while almost all potato isolates belong to the O or N groups (Blanco-Urgoiti et al., 1998). Several recombinant isolates have also been characterized, mainly between the N and O groups (Revers et al., 1996; Glais et al., 2002; Ogawa et al., 2008). Two recessive resistance alleles at the pvr2 locus have been used extensively for more than 50 years to control PVY in pepper crops. The pvr2 allele is highly durable since no resistance breakdown has been observed in the field and very few isolates have been described that are able to infect plants carrying that resistance (Gébré-Sélassié et al., 1985; Luis Arteaga et al. 1993). The pvr2 allele is less durable since it has been broken down in some regions. However, V er si on p os tp rin t Comment citer ce document : Ben Khalifa, M., Simon, V., Fakhfakh, H., Moury, B. (2012). Tunisian Potato virus Y isolates with unnecessary pathogenicity towards pepper : Support for the matching allele model in eIF4E resistance – potyvirus interactions. Plant Pathology, 0032-0862, 61 (3), 441-447. DOI : 10.1111/j.1365-3059.2011.02540.x Version définitive du manuscrit publié dans / Final version of the manuscript published in : Plant Pathology, 2012, 61, 441-447; DOI: 10.1111/j.1365-3059.2011.02540.x. The original publication is available at http://onlinelibrary.wiley.com/doi/10.1111/j.13653059.2011.02540.x/abstract. M an us cr it d’ au te ur / A ut ho r m an us cr ip t M an us cr it d’ au te ur / A ut ho r m an us cr ip t M an us cr it d’ au te ur / A ut ho r m an us cr ip t pvr2-breaking PVY isolates are usually less frequent than avirulent ones (Luis Arteaga and Gil Ortega 1986). Seven additional pvr2 alleles confer particular resistance specificity towards PVY isolates and correspond to particular amino acid substitutions in the pvr2-encoded product, i.e. the eukaryotic translation initiation factor 4E (eIF4E) (Charron et al. 2008). These alleles have not been used in the breeding of elite F1 hybrids but are present in local pepper populations. Interaction of PVY isolates with the four pvr2 alleles pvr2 (the susceptibility allele), pvr2, pvr2 and pvr2 is controlled by a 23-amino-acid-long region in the viral genome-linked protein (VPg) (Ayme et al., 2006, 2007). Amino acid substitutions in VPg and eIF4E can modify their binding properties and are responsible for the pepper resistance specificity and PVY pathogenicity. When this interaction is impaired, the virus fails to accumulate in inoculated tissues and to complete its infection cycle (Charron et al. 2008). Genetic diversity of PVY populations infecting pepper in northern Tunisia was analysed by RFLP haplotypes of three genome regions, including the VPg-coding region (Ben Khalifa et al., 2009). The study revealed a high genetic diversity and a strong genetic differentiation between the PVY populations due to sampling locality. The sampled plants were either susceptible to PVY (they carried the susceptibility allele pvr2) or homozygous for the pvr2 allele and hence potentially resistant to some PVY isolates. Surprisingly, the polymorphism at the pvr2 locus of the sampled plants did not contribute significantly to the genetic differentiation of PVY. The aim of this study was consequently to characterize the pathogenicity of some of these Tunisian PVY isolates towards the pepper resistances at the pvr2 locus. MATERIALS AND METHODS PVY isolates Details of all PVY isolates used in this study, their locality of origin, collection year, original host plant and, for some of them, the pvr2 genotype of these plants together with GenBank accession numbers are shown in table 1. Tunisian PVY isolates used in this study were obtained from fieldinfected Capsicum annuum plants in northern Tunisia in 2005 or 2006 or in 1994 in the centre of Tunisia (region of Kairouan) or in northern Tunisia (Borj el Amri). Isolates from 2006 were chosen to represent different haplotypes previously described by Ben Khalifa et al. (2009) and to avoid plants infected simultaneously by several virus variants. Other isolates were chosen randomly. Since their isolation, they were propagated once onto the laboratory host Nicotiana tabacum cv. Xanthi-nc and were stored as dehydrated infected leaves. cDNA synthesis and sequencing of the PVY VPg cistron Total RNAs were purified from a 0·5 g piece of flesh of PVY-infected pepper fruits with the TriReagent kit (Molecular Research Center Inc.), and were used as template for RT-PCR (Moury et al., 2004). Part of the VPg cistron was amplified with primers VPg-F and VPg-R (Ben Khalifa et al., 2009) designed to be polyvalent for all PVY groups and amplified 505 nucleotides of the VPg cistron. Sequencing reactions were performed directly on RT-PCR products with primer VPg-F by Genome Express (Grenoble, France). Phylogenetic analyses Using the CLUSTAL W program (Thomson et al., 1994), we obtained a 462-nucleotide-long alignment of the PVY VPg coding region, corresponding to the region from position 5796 to position 6257 of the genome of isolate SON41p (accession number AJ439544). Phylogeny construction and evaluation was done using the neighbor-joining (NJ) method implemented in the MEGA software (Tamura et al., 2007) and the robustness of the tree topology was evaluated with 1,000 bootstrap resampling. The codeml program of the software PAML version 4.2 (Yang 2007) was used to estimate the selection intensity in the VPg central region, which is involved in PVY pathogenicity towards the pvr2 resistances (codon positions 101 to 123), along the different branches of the tree using the majority-rule consensus tree topology obtained above. For this, estimates of the numbers of synonymous (s) and nonsynonymous (n) nucleotide substitutions were obtained for each branch of V er si on p os tp rin t Comment citer ce document : Ben Khalifa, M., Simon, V., Fakhfakh, H., Moury, B. (2012). Tunisian Potato virus Y isolates with unnecessary pathogenicity towards pepper : Support for the matching allele model in eIF4E resistance – potyvirus interactions. Plant Pathology, 0032-0862, 61 (3), 441-447. DOI : 10.1111/j.1365-3059.2011.02540.x Version définitive du manuscrit publié dans / Final version of the manuscript published in : Plant Pathology, 2012, 61, 441-447; DOI: 10.1111/j.1365-3059.2011.02540.x. The original publication is available at http://onlinelibrary.wiley.com/doi/10.1111/j.13653059.2011.02540.x/abstract. M an us cr it d’ au te ur / A ut ho r m an us cr ip t M an us cr it d’ au te ur / A ut ho r m an us cr ip t M an us cr it d’ au te ur / A ut ho r m an us cr ip t the tree and estimates of the numbers of synonymous (S) and nonsynonymous (N) sites were obtained for the sequence alignment. A test of positive selection and tests of episodic evolution along a specific branch of the tree were performed as in Zhang et al. (1997) (see also Table 2). For the test of positive selection, the numbers of synonymous and nonsynonymous substitutions estimated to have occurred along that particular branch were compared to the expected numbers under the null hypothesis of neutral evolution. For the tests of episodic evolution, the numbers of synonymous and nonsynonymous substitutions estimated to have occurred along that particular branch were compared to the sum of the numbers of synonymous and nonsynonymous substitutions estimated to have occurred (i) in all branches corresponding to the descendant lineages or (ii) in all branches corresponding to the lineages of other clades. Fisher exact tests of homogeneity were performed for all these comparisons. Biological characterization of PVY pathotypes The C. annuum inbred lines used for characterization of the pathogenicity properties of PVY isolates were Yolo Wonder (pvr2/pvr2), Yolo Y (pvr2/pvr2), Florida VR2 (pvr2/pvr2) and HD285 (pvr2/pvr2). Virus isolates kept as dehydrated material were first multiplied on Xanthi-nc plants and PVY-infected leaves of these plants were used to prepare inocula for pepper genotypes. For each PVY isolate and each pepper genotype, 20 seedlings at the two-cotyledon stage (two to three weeks after sowing) were inoculated on the two cotyledons as in Moury et al. (2004). One month post inoculation, PVY infection of the pepper plants was assessed by DAS-ELISA (Moury et al. 2004). RESULTS Diversity of the VPg pathogenicity factor of Tunisian PVY isolates from pepper About 100 PVY isolates have been isolated from pepper plants in Northern Tunisia in 2006 and previously subjected to an RFLP analysis covering three genome regions (Ben Khalifa et al. 2009). This study revealed a high level of genetic variability and a large number of plants infected by mixtures of different PVY haplotypes. On this basis, we chose 13 isolates that belonged to different haplotypes and that did not show evidence of mixed infection. We also analysed nine isolates collected in 2005 and chosen randomly among a collection of 50 isolates, and five isolates collected in 1994 in Northern or central Tunisia (Table 1). A 462 nucleotide long sequence of the VPg coding region (from position 5796 to position 6257 of the reference genome of isolate SON41p; accession number AJ439544) was obtained for all these isolates (GenBank accession numbers JF824713 to JF824736). There were very few ambiguous nucleotides (double peaks) in the sequence chromatograms and in these cases there was usually a marked difference in the peak heights. At these positions, only the predominant nucleotide in the population was considered in further analyses. Phylogenetic analysis of the sequenced region revealed three separate groups of isolates supported by high bootstrap values (>99%; Fig. 1). Compared to the overall PVY diversity, these isolates belonged to the C1 subgroup of clade C, as do most pepper isolates (96% bootstrap support; Blanco-Urgoiti et al. 1998 and data not shown). These three groups possess particular amino acids in the central part of the VPg (amino acid positions 101 to 123) which was shown to determine the pathogenicity properties of PVY towards the recessive resistances at the pvr2 locus in pepper (Ayme et al. 2006, 2007) (Fig. 2). In this VPg region, all group 1 isolates possess identical amino acids with the exception of isolate K1694 which is composed of a mixture of viruses possessing either a valine or a leucine at position 118 (Fig. 2). Similarly, all group 3 isolates possess identical amino acids in this region. Finally, group 2 isolates show two different amino acid sequences that differ at position 115 only (Fig. 2). The fact that the central part of PVY VPg is subjected to positive selection (Moury et al. 2004) could induce a biased image of the phylogenetic relationships between PVY isolates. Consequently, we also performed the phylogenetic analysis after excluding codon positions 101 to 123 from the sequence alignment. This second analysis revealed the same tree topology as before, the three major PVY groups showing high bootstrap support (>90%; data not shown). The tree topology was consequently V er si on p os tp rin t Comment citer ce document : Ben Khalifa, M., Simon, V., Fakhfakh, H., Moury, B. (2012). Tunisian Potato virus Y isolates with unnecessary pathogenicity towards pepper : Support for the matching allele model in eIF4E resistance – potyvirus interactions. Plant Pathology, 0032-0862, 61 (3), 441-447. DOI : 10.1111/j.1365-3059.2011.02540.x Version définitive du manuscrit publié dans / Final version of the manuscript published in : Plant Pathology, 2012, 61, 441-447; DOI: 10.1111/j.1365-3059.2011.02540.x. The original publication is available at http://onlinelibrary.wiley.com/doi/10.1111/j.13653059.2011.02540.x/abstract. M an us cr it d’ au te ur / A ut ho r m an us cr ip t M an us cr it d’ au te ur / A ut ho r m an us cr ip t M an us cr it d’ au te ur / A ut ho r m an us cr ip t reliable and was not mainly influenced by a small number of positively-selected amino acid sites. There was no link between the PVY groups and the geographic or cultivar origin of the isolates (Table 1). The genotype at the pvr2 locus of the sampled plants was determined previously for the isolates collected in 2006. Two groups of plants can be distinguished: susceptible plants that carry the pvr2 allele and pvr2/pvr2 plants that are potentially resistant to some PVY isolates (Ben Khalifa et al. 2009). All of four isolates collected on pvr2/pvr2 plants belonged to group 1 of isolates, while the nine isolates collected on pvr2/plants were shared between the three PVY groups. Despite the small number of isolates, this discrepancy suggested some correspondence between the phylogeny and pathogenicity properties of the isolates. Pathogenicity of Tunisian PVY isolates against pepper genotypes carrying different pvr2 alleles The pathogenicity of nine isolates belonging to the different PVY groups was evaluated against a series of reference pepper genotypes homozygous for the pvr2, pvr2, pvr2 or pvr2 alleles (Table 1). Pathogenicity could not be established by the occurrence of symptoms on inoculated plants because all isolates were in mixture with Cucumber mosaic virus (CMV; genus Cucumovirus, family Bromoviridae). Consequently, DAS-ELISA detection of PVY was performed one month post-inoculation on apical leaves. On this basis, two categories of PVY isolates were observed: isolates that infected all plants from the reference susceptible genotype (pvr2/pvr2) only and isolates that infected all plants except those from the pvr2/pvr2 genotype (Table 1). According to the nomenclature for PVY pathotypes (for example Ayme et al. 2007), the former isolates belonged to pathotype (0) and the latter to pathotype (0,1,3). The absence of infection of pepper plants carrying the pvr2, pvr2 or pvr2 resistances by a large number of PVY isolates indicated the absence of synergism between CMV and PVY in regard with these eIF4E-mediated resistances and validated the analysis of the pathogenicity of the PVY isolates. There was a perfect correspondence between the pathogenicity properties and phylogeny. PVY isolates from group 1 belonged to pathotype (0,1,3) while isolates from groups 2 and 3 belonged to pathotype (0). This was logical considering that the phylogeny itself corresponded to the polymorphism observed in the central part of the VPg. Phylogenetic analysis of selection intensity in PVY VPg Given the correspondence between phylogeny and pathogenicity properties of Tunisian PVY isolates, we asked whether the diversification into three major groups could have been the result of the selection imposed by the host plant resistance. For this, the PAML software was used to estimate the selection intensity within the region involved in pathogenicity (i.e. codon positions 101 to 123) and along the different branches of the phylogenetic tree obtained previously (Fig. 1). A single branch showed an apparent excess of nonsynonymous substitutions: the branch linking PVY group 1, corresponding to pathotype (0,1,3), to the rest of the tree. Along that branch, 8.1 nonsynonymous substitutions and zero synonymous substitutions were estimated to have occurred (Fig. 1). A formal test of positive selection comparing observed and expected synonymous and nonsynonymous changes along that particular branch (Zhang et al. 1997) was however only marginally significant (P-value=0.10, Fisher exact test; Table 2A), but the small number of substitutions does not provide a high statistical power. Tests of episodic evolution, comparing the selection intensity along that particular branch and either (i) along the set of branches linking the descendant lineages in group 1 or (ii) along the set of branches corresponding to groups 2 and 3 were highly significant (P-values=0.0018 and 0.001, respectively; Fisher exact tests; Tables 2B and 2C). DISCUSSION Evolution towards pathogenicity against the pvr2 resistance allele in Tunisian PVY isolates The two isolates from group 1 whose pathogenicity properties were evaluated against the pvr2 alleles (isolates AOUA2 and GHD4) belonged to pathotype (0,1,3). Given that all isolates from group 1 share identical amino acid residues at VPg positions 101 to 123 and that this region is the V er si on p os tp rin t Comment citer ce document : Ben Khalifa, M., Simon, V., Fakhfakh, H., Moury, B. (2012). Tunisian Potato virus Y isolates with unnecessary pathogenicity towards pepper : Support for the matching allele model in eIF4E resistance – potyvirus interactions. Plant Pathology, 0032-0862, 61 (3), 441-447. DOI : 10.1111/j.1365-3059.2011.02540.x Version définitive du manuscrit publié dans / Final version of the manuscript published in : Plant Pathology, 2012, 61, 441-447; DOI: 10.1111/j.1365-3059.2011.02540.x. The original publication is available at http://onlinelibrary.wiley.com/doi/10.1111/j.13653059.2011.02540.x/abstract. M an us cr it d’ au te ur / A ut ho r m an us cr ip t M an us cr it d’ au te ur / A ut ho r m an us cr ip t M an us cr it d’ au te ur / A ut ho r m an us cr ip t sole pathogenicity determinant towards the pvr2 resistance alleles in pepper (Ayme et al. 2006, 2007), we inferred that all group 1 isolates belonged to pathotype (0,1,3). For similar reasons, we considered that all isolates from groups 2 and 3 belonged to pathotype (0). As a consequence, the most parsimonious scenario for acquisition of pathogenicity toward pvr2 and pvr2 corresponds to a single event (along branch (a); Fig. 1). This event was probably the result of the selection imposed by resistant pepper plants, though the positive selection test along that particular branch of the phylogenetic tree was only marginally significant (Table 2A). Indeed, a significant acceleration of amino acid substitutions in the VPg region determining pathogenicity properties was observed along that branch compared to the rest of the tree (Tables 2B and 2C). Since about eight amino acid substitutions were estimated to have occurred along that branch, it is probable that several of them were required for resistance breakdown. Each of two amino acid substitutions observed in the VPg of isolates from group 1, the serine to glycine substitution at position 101 and the threonine to arginine substitution at position 115, have been described as sufficient for breakdown of the pvr2 resistance (Ayme et al. 2006). The other amino acid substitutions are specific to this group of Tunisian isolates and have not been shown to alter PVY pathogenicity before. If several amino acid substitutions were required for acquisition of resistance breaking properties by group 1 isolates, this could explain why such event was rare during PVY history and why isolates from the two other groups did not succeed to break the pvr2 resistance of Tunisian pepper cultivars down. This event is not very recent since isolate K1694, collected in 1994, belongs also to group 1. It is possible that emergence of group 1 occurred locally. No other PVY isolate close to group 1 is present in sequence databanks, while isolate LYE84.2 from the Canary Islands (accession number AJ439545) belongs to group 2 and French isolates To72 and SON41p (accession numbers EU334782 and AJ439544, respectively) are close to group 3 (Fig. 1). Choosing the most efficient pvr2 resistance allele to control PVY in pepper in Tunisia Quite a large number of amino acid substitutions in PVY VPg have been identified that confer particular pathogenicity properties towards the pvr2 resistance alleles of pepper (Ayme et al. 2006, 2007). These data can be used to choose the most appropriate resistance allele in order to control PVY in a given geographical region. The most widespread resistance allele in the sampled Tunisian pepper cultivars is pvr2 (Ben Khalifa et al. 2009) and pvr2-resistance breaking isolates are widespread in these crops (Table 1, Fig. 2). Although pvr2 is not present in the sampled Tunisian pepper cultivars, isolates from group 1 are already able to break this resistance down. Consequently, the pvr2 resistance will probably be useless to control PVY in these crops. The resistance conferred by pvr2 would certainly be more appropriate. This resistance has been used worldwide for ca. 50 years and has not been broken down. Only two PVY isolates able to infect pvr2/pvr2 plants have been described. The SON41p isolate has been selected from a pathotype (0,1) isolate by serial passages in pvr2/pvr2 pepper plants (Gebre Selassie et al. 1985). The second case is a field isolate collected in the region of Málaga, Spain (Luis Arteaga et al. 1993) that was not mentioned afterwards. This latter isolate was not characterized at the molecular level. Isolates that are closest to SON41p at the critical residues in the centre of the VPg that control PVY pathogenicity properties towards the pvr2 alleles differ by two amino acids (corresponding to two nucleotide substitutions in the coding region) and are similar to isolate To72 (accession number EU334782; Figs. 2 and 3) which belongs to pathotype (0) (Ayme et al. 2007). The combination of these two substitutions is required for infection of pvr2/pvr2 plants (Ayme et al. 2007). Tunisian isolates from group 3 also differ by two amino acids from SON41p and evolution towards a VPg identical to that of SON41p would require three nucleotide substitutions (and three amino acid substitutions) (Figs. 2 and 3), which is less likely than for To72. Isolates from other groups are more distant from SON41p and consequently much less likely to break the pvr2 resistance down. Consequently, the pvr2 resistance would certainly confer an efficient and durable resistance towards the Tunisian PVY pepper isolates. The apparent lack of fitness cost associated to pathogenicity against the pvr2 resistance alleles supports the matching allele model for plant-virus interactions V er si on p os tp rin t Comment citer ce document : Ben Khalifa, M., Simon, V., Fakhfakh, H., Moury, B. (2012). Tunisian Potato virus Y isolates with unnecessary pathogenicity towards pepper : Support for the matching allele model in eIF4E resistance – potyvirus interactions. Plant Pathology, 0032-0862, 61 (3), 441-447. DOI : 10.1111/j.1365-3059.2011.02540.x Version définitive du manuscrit publié dans / Final version of the manuscript published in : Plant Pathology, 2012, 61, 441-447; DOI: 10.1111/j.1365-3059.2011.02540.x. The original publication is available at http://onlinelibrary.wiley.com/doi/10.1111/j.13653059.2011.02540.x/abstract. M an us cr it d’ au te ur / A ut ho r m an us cr ip t M an us cr it d’ au te ur / A ut ho r m an us cr ip t M an us cr it d’ au te ur / A ut ho r m an us cr ip t Sacristán and García-Arenal (2008) proposed that the interaction and coevolution between plant recessive resistances and viruses correspond to the “matching allele” model. Under that model, “infection of plants requires a specific match between host and parasite genes” (Sacristán and García-Arenal 2008). For the pepper-PVY interaction, this was established by Charron et al. (2008) who showed that infection depended on the direct binding between the plant eIF4E and virus VPg, and that amino acid substitutions in pepper eIF4E or PVY VPg that disrupt binding conferred an absence of infection of the plant. At the population scale, the matching allele model predicts (i) that no parasite genotype evolves a general pathogenicity on all host genotypes and (ii) that there is no fitness cost associated to pathogenicity (Agrawal and Lively 2002). The first prediction was partly validated in the PVY-pepper interaction since a large number of PVY mutants that gained pathogenicity toward a particular pvr2 allele simultaneously lost their pathogenicity towards other alleles (Ayme et al. 2007). Concerning the second prediction, no exhaustive fitness comparisons have been made in controlled laboratory conditions between PVY isolates with different pathogenicity properties. However, PVY epidemiological data in Tunisia suggest lack of, or low fitness costs associated to pathogenicity toward the pvr2 resistances. First, there was no effect of the pvr2 resistance in the sampled plants on the genetic structure of PVY populations (Ben Khalifa et al. 2009). This was probably due to the fact that a large number of pathotype (0,1,3) isolates are present in pvr2/pvr2 plants. In the sampled areas, an average of 23% of plants possess a pvr2/pvr2 genotype (Ben Khalifa et al. 2009). If pathotype (0,1,3) isolates were as fit as pathotype (0) isolates in susceptible pvr2/plants, their expected frequency would be 61.5% on average (the 23% of pvr2/pvr2 plants plus half of the 77% of pvr2/pvr2 plants would be infected by pathotype (0,1,3) isolates while half of the 77% of pvr2/pvr2 plants would be infected by pathotype (0) isolates; 1×0.23+0.5×0.77=0.615). Twelve of 22 PVY isolates (54.5%) belong to pathotype (0,1,3) in our set of isolates either chosen randomly (those from 2005) or at random among RFLP haplotypes (those from 2006) (Table 1). This observed value is not significantly different from expectations under an equal-fitness model (the 95% confidence interval is [8, 18] for pathotype (0,1,3) isolates under this model; Monte Carlo simulations with the R software). Consequently, there is no apparent fitness cost of pathotype (0,1,3) isolates in susceptible pvr2/pepper plants in the Tunisian epidemiological conditions. Isolates from group 1 are not only infectious in pvr2/pvr2 plants but they can also infect pvr2/pvr2 plants, although this genotype is absent or very rare in Tunisia. Consequently, pathogenicity towards pvr2 also does not seem to be costly for group 1 isolates. It is plausible that pathogenicity towards pvr2 is a byproduct of pathogenicity selected by the pvr2 resistance due to the fact that the three-dimensional structure of the VPg of group 1 isolates was by chance able to bind the eIF4Es encoded by both pvr2 and pvr2. Then, these isolates were maintained in PVY populations by the selective pressure of pvr2/pvr2 plants and the absence of (or low) counter selection in pvr2/plants. Acknowledgements This study was supported by a Franco-Tunisian CORUS program from the French Ministère des Affaires Etrangères, the Tunisian Ministère de l’Education Supérieure, de la Recherche Scientifique et de la Technologie, a program ‘Ecologie pour la gestion des Ecosystèmes et de leurs Ressources’ and a program ‘Agriculture et Développement Durable’ from the Agence Nationale de la Recherche. We thank Julie Quenouille for interesting comments on an earlier version of the