Microsoft Word - 2011-moury-mbe

The dN/dS ratio between nonsynonymous and synonymous substitution rates has been used extensively to identify codon positions involved in adaptive processes. However, the accuracy of this approach has been questioned and very few studies have attempted to validate experimentally its predictions. Using the coat protein (CP) of Potato virus Y (PVY; genus Potyvirus, family Potyviridae) as a case study, we identified several candidate positively-selected codon positions which differed between clades. In the CP of the N clade of PVY, positive selection was detected at codon positions 25 and 68 by both the softwares PAML and HyPhy. We introduced nonsynonymous substitutions at these positions in an infectious cDNA clone of PVY and measured the effect of these mutations on virus accumulation in its two major cultivated hosts, tobacco and potato, and on its efficiency of transmission from plant to plant by aphid vectors. The mutation at codon position 25 significantly modified the virus accumulation in the two hosts while the mutation at codon position 68 significantly modified the virus accumulation in one of its hosts and its transmissibility by aphids. Both mutations were involved in adaptive trade-offs. We suggest that our study was particularly favourable to the detection of adaptive mutations using dN/dS estimates because, as obligate parasites, viruses undergo a continuous and dynamic interaction with their hosts which favours the recurrent selection of adaptive mutations and because trade-offs between different fitness traits impede (or at least slow down) the fixation of these mutations and maintain polymorphism within populations. Version définitive du manuscrit publié dans / Final version of the manuscript published in : Molecular Biology and Evolution, 2011, 28 (9), 2707-2717 DOI: 10.1093/molbev/msr105. The publisher version is available at http://mbe.oxfordjournals.org/content/28/9/2707.full.pdf+html. 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 Introduction Since the beginning of the years 2000, methods based on estimation of the ratio dN/dS between the nonsynonymous (dN) and the synonymous (dS) substitution rates in an alignment of amino-acidcoding sequences (named “dN/dS methods” in the rest of the article) have been used extensively to identify individual codon positions evolving under positive selection (Nielsen 2005). However, both the theoretical grounds of these methods and their accuracy to detect adaptive codon substitutions have been matters of concern (for example Suzuki and Nei 2004; Wyckoff et al. 2005; Galtier and Duret 2007; Hughes 2007, 2008; Kryazhimskiy and Plotkin 2008; Nozawa et al. 2009). Although highly desirable, experimental validation (or refutation) of these predicted positively-selected codon positions is presently extremely scarce. In their study of vertebrate vision genes, Yokoyama et al. (2008) surprisingly found no correspondence between functionally-important amino acid replacements during vertebrate evolution and amino acid positions under positive selection according to dN/dS methods. Here, we used dN/dS methods to detect candidate amino acid positions for positive selection in a viral protein and we validated these predictions by measuring the effect of nonsynonymous substitutions at these positions on the virus fitness. For this, we focussed on Potato virus Y (PVY) (genus Potyvirus, family Potyviridae), a major pathogen of crops belonging to the family Solanaceae, including potato, tobacco and pepper. PVY genome consists of a single-stranded positive-sense RNA which encodes a polyprotein further matured into eleven functional proteins. A previous study suggested that 12% of amino acid positions in the PVY coat protein (CP) were evolving under slight positive selection (Moury et al. 2002). However, the small number of sequences available at that time did not allow a more precise clade-by-clade analysis. More than 100 PVY CP sequences are now available, making a thorough analysis of the selection intensity in this protein feasible. PVY is transmitted horizontally from plant to plant by more than 40 aphid species (family Aphididae) in a nonpersistent manner (Edwardson and Christie 1997) where the virus is retained only transiently by the vector. Vertical transmission of PVY from infected plants to next generation through tubers is also a major problem for the production of commercial seed potatoes. The CP of potyviruses is involved in indirect binding of the virions to the aphid mouthparts (Peng et al. 1998; Blanc et al. 1997) and in plant infection (more specifically in virus translocation within plants; Dolja et al. 1994, 1995; Andersen and Johansen 1998; Atreya et al. 1995; López-Moya and Pirone 1998) and can therefore be a target of selection by both host plants and vectors. Materials and Methods Data collection and phylogenetic analyses Sequences of the PVY CP coding region were downloaded from GenBank in December 2008. Redundant sequences and sequences containing nucleotide ambiguities were discarded from the dataset. Sequences were aligned using the CLUSTALW program (Thompson et al., 1994). The sequence alignment was analysed with RDP version 2 software (Martin et al., 2005) implementing several algorithms to detect recombinant sequences. Phylogenetic methods were used to determine the branching pattern of PVY sequences and to perform analysis of the selection intensity along the CP. The maximum likelihood (ML) method implemented in PhyML version 3.0 (Guindon and Gascuel 2003), incorporating the best-fit nucleotide substitution model estimated by MODELTEST (Posada and Crandall, 1998), and the neighbour joining (NJ) method implemented in MEGA version 4 software (Tamura et al. 2007) were used to infer the tree topology. Bootstrap resampling was performed to assess branch support. To identify codon sites with positive selection, i.e. sites showing a dN/dS ratio significantly higher than one (Kimura, 1983), ML approaches implemented in the HyPhy (Pond and Frost, 2005a) and in the PAML version 4.2 (Yang 2007) packages were performed using the majority-rule consensus tree topology obtained with the NJ method. In HyPhy, the SLAC (Single-Likelihood Ancestor Counting), FEL (fixed effects likelihood) and REL (random effects likelihood) methods (Pond and Frost 2005a, 2005b; Pond et al. 2006) were used. Under PAML, the nested evolution models M0, Version définitive du manuscrit publié dans / Final version of the manuscript published in : Molecular Biology and Evolution, 2011, 28 (9), 2707-2717 DOI: 10.1093/molbev/msr105. The publisher version is available at http://mbe.oxfordjournals.org/content/28/9/2707.full.pdf+html. 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 M1a, M2a and M3 from one part, and the nested models M7 and M8 from the other part were compared using likelihood ratio tests (LRTs) to assess the significance of positive selection in the sequence alignments and empirical Bayes methods allowed identifying individual positivelyselected codon sites a posteriori (Yang et al. 2000; Yang 2007). Each PAML model was run at least three times with different initial values for dN/dS to avoid local maximum likelihood estimates. The LRT developed by Knudsen and Miyamoto (2001) was used to detect specific nucleotide sites that evolve at different rates in the different PVY clades. A significant rate difference between two subgroups at a given site would, thereby, mean that the function of this position could be different in the two clades and/or that evolutionary constraints differ between PVY clades. The likelihood of the null hypothesis assuming that a given position evolves with different rates in the two sequence groups is compared with the likelihood of the alternative hypothesis (same rate in the two subgroups). The program is available at http://www.daimi.au.dk/~compbio/rateshift/ and allows analysis of 30 sequences of a maximum length of 600 nucleotides at the same time. Consequently, two random subsets of 30 sequences encompassing nucleotide positions 1 to 600 or 202 to 801 were analyzed for each clade. Mutagenesis of an infectious PVY clone An infectious cDNA clone of PVY isolate N605, a non-recombinant isolate of clade N, was built previously (Jakab et al. 1997; Bukovinszki et al. 2007). This clone was modified to include a cassette containing the 2μ yeast replication origin and a selectable marker (Trp-1 promoter and gene) as described in Fernandez-Delmond et al. (2004) and Ayme et al. (2006), allowing homologous recombination in yeast. A cDNA clone corresponding to the CP-coding region of N605 was mutated at nucleotide position 74 (A74T, i.e. adenosine to thymidine substitution changing the asparagine at position 25 of the CP to an isoleucine) or at nucleotide position 202 (G202A, i.e. guanosine to adenosine substitution changing the glutamic acid at position 68 of the CP to a lysine) using the QuikChange site-directed mutagenesis kit (Stratagene, La Jolla, CA). The region coding the CP was deleted by reverse PCR from the N605 clone and a unique NotI restriction site was introduced at the junction, yielding the acceptor shuttle vector. Yeast was cotransformed with the CP-deleted vector linearized with NotI and with cDNA fragments of the CP coding region containing the nucleotide substitution which yielded through homologous recombination the mutated N605 clones N605-CP25 and N605-CP68. The conformity of the clones was checked by sequencing to rule out spurious mutations. Primary inoculations with the cDNA clones were made by DNA-coated tungsten particle bombardments of Nicotiana clevelandii juvenile plants (4 weeks post sowing), because direct bombardment of potato or tobacco plants was not efficient. Comparison of accumulation of PVY N605 wild-type clone and mutants in tobacco and potato plants The fitness of the N605-CP25 or N605-CP68 mutant clones was compared to that of the parental N605 clone in both single-inoculation and competition experiments in tobacco (Nicotiana tabacum cv. Xanthi) and potato (Solanum tuberosum cv. Bintje). Inocula were calibrated using DAS-ELISA as shown by Ayme et al. (2006) and used separately for single inoculations or mixed for competition experiments. For single inoculations, the relative accumulation of each variant was measured at 30 days post inoculation (dpi) in plants inoculated either by N605, N605-CP25 or N605-CP68. For each plant, all leaves were pooled, ground with buffer (1:4 wt/vol), and fivefold successive dilutions in buffer of each plant extract were tested by DAS-ELISA as in Ayme et al. (2006). The relative PVY accumulation in these samples was determined using absorbance values measured at 405 nm (A405) where the curves representing A405 in function of the sample dilution factor were linear and parallel between samples (Ayme et al. 2006). Since the two amino acid substitutions introduced in the CP of the N605 clone could modify their reactivity towards the polyclonal antibodies used for DASELISA and could consequently introduce a bias in our estimation of the relative concentration of the different PVY variants in plants, we measured the reactivity of these variants as above, using Version définitive du manuscrit publié dans / Final version of the manuscript published in : Molecular Biology and Evolution, 2011, 28 (9), 2707-2717 DOI: 10.1093/molbev/msr105. The publisher version is available at http://mbe.oxfordjournals.org/content/28/9/2707.full.pdf+html. 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 DAS-ELISA with several dilutions of virus of known absolute concentration purified from Xanthi plants at 10 to 14 dpi and quantified (Moury et al. 2007). Comparison of DAS-ELISA reactivity of the PVY variants was done with the results obtained from four independent virus purifications with a Kruskal-Wallis nonparametric test. A Kruskal-Wallis test was also performed on the relative accumulation of each PVY variant, with and without correction for relative DAS-ELISA reactivity, in order to detect significant accumulation differences between the PVY variants. For competition experiments, inocula were mixed at a 1:1 ratio based on DAS-ELISA and used to inoculate plants. Calibration of the relative concentration of inocula using this method was shown to be similar to a calibration using a quantitative RT-PCR method (Janzac et al. 2010). The competitiveness of the PVY variants was analyzed using dCAPS analysis (Michaels and Amasino 1998; Neff et al. 1998) at 30 dpi. For each inoculated plant, total RNAs were extracted from pools of three infected leaves with the Tri Reagent kit (Molecular Research Center Inc., Cincinnati, OH) and used to amplify part of the CP coding region by RT-PCR with primers ‘dCAPS-N25I’ and ‘Reverse’ or ‘dCAPS-E68K’ and ‘Reverse’ (Supplementary Table S1), depending on the competition. The presence of the mutations A74T and G202A in the virus populations was assessed with the endonucleases EcoRI and MseI, respectively, which cleave RT-PCR products obtained with N605-CP25 or N605-CP68, but not with N605. Plants were classified into three categories depending on the observation of cleaved or uncleaved RT-PCR products, or both, after 2% agarose gel electrophoresis. Fisher exact tests were performed to compare the competitiveness of the virus variants and to analyse differential selection by the two plant species. For this, the mixed-infected plants were shared between the sets of plants infected by the wild-type or the mutant virus only. Then, to compare the virus competitiveness, the observed number of plants infected by the wildtype or the mutant virus was compared to the numbers expected if the variants had equal fitnesses, i.e. to the 1:1 distribution corresponding to the inoculum ratio. Differential selection by host species was assessed for each virus competition by comparing the observed numbers of plants infected by the wild-type or the mutant virus at 30 dpi in tobacco and potato. The sequence of the CP coding region of the progeny of the cloned PVY mutants in 10 randomlychosen inoculated tobacco and potato plants was determined to exclude that second-site mutations in this region had modified the fitness properties of the virus or had hindered dCAPS analysis. The infectivity of N605 and the two derived CP mutants was evaluated by manual inoculation of a series of diluted virus solutions prepared from calibrated inocula to 20 Xanthi plants per dilution and analysis of the frequency of infected plants as assessed by DAS-ELISA. Comparison of the transmission efficiency of PVY N605 clone and mutants by aphids Wingless aphids of the species Myzus persicae Sulzer or Aphis gossypii Glover were starved during 3 to 4 hours prior acquisition. Sets of ten aphids were allowed an acquisition access period of 2 min ± 30 sec on PVY-infected leaves of the tobacco cultivar Xanthi. Each tobacco leaf was used as virus source for a single set of ten aphids. Aphids were then transferred with a paintbrush to healthy Xanthi seedlings at the two-expanded-leaf stage (one aphid per test plant). Aphids were maintained on plants overnight and treated with an insecticide before being transferred to an insect-proof greenhouse for one month. The infection status of inoculated plants was then checked by symptom reading and DAS-ELISA on apical, uninoculated leaves. The relative concentration of PVY in the Xanthi leaves that were used for virus acquisition by aphids was determined by DAS-ELISA as described above. Results The pattern of selection in the CP-coding region of PVY varies across clades After discarding redundant sequences and sequences containing ambiguous nucleotides, a dataset of 125 sequences of the CP coding region (801-nucleotide long) was obtained and aligned. No gaps were found in the alignment. Recombination in the CP-coding region of PVY is well documented (Revers et al. 1996 Ogawa et al. 2008) and datasets containing recombinant sequences can provide false positives in selection analyses (Anisimova et al. 2003; Shriner et al. 2003). Our purpose was to measure the selection intensity in the PVY CP in a clade-by-clade manner. Consequently, we Version définitive du manuscrit publié dans / Final version of the manuscript published in : Molecular Biology and Evolution, 2011, 28 (9), 2707-2717 DOI: 10.1093/molbev/msr105. The publisher version is available at http://mbe.oxfordjournals.org/content/28/9/2707.full.pdf+html. 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 discarded clades that contained less than 10 sequences, since a reliable estimation of selection intensity was not possible with so few sequences. This provided a dataset composed of 119 sequences (Supplementary Material) for which the phylogenetic trees generated with PhyML or MEGA were consistent, had high bootstrap support at major branches and separated PVY isolates among four clades (Fig. 1), including three nonrecombinant clades (named clades C, N and O; Moury et al. 2010) and a clade of recombinant isolates showing a recombination breakpoint between nucleotide positions 603 and 616 (Revers et al. 1996) of the CP coding region (named “recombinant clade” in the following). Members of this recombinant clade show close similarity with isolates of clades N and O for the 5’ and 3’ parts of the CP coding region, respectively. Recombination was validated with RDP and by separate phylogenetic analyses of the two CP coding regions. When this recombinant clade was withdrawn, no more recombination signal was detected with RDP. The selection intensity was analysed separately for the four different clades and clade N was analysed together with the recombinant clade for the region spanning nucleotides 1 to 600, upstream of the recombination breakpoint (Table 1). Using PAML, no codon site under positive selection was detected for clade C, while significant positive selection was detected for amino acid positions 10, 11 and 26 for clade O and for amino acid positions 25 and 68 in the alignment of the 600 nucleotides at the 5’ extremity of the CP coding region of the set of isolates belonging to clade N or to the recombinant clade (Table 1). The dN/dS ratio of the positively-selected category of codons was estimated to be 5.6 for the N (and recombinant) clade and 6.7 for clade O (Table 1). These candidate codon positions for positive selection were confirmed with the REL method implemented in HyPhy (Table 1). This method detected also codon positions 26 and 139 for the N (and recombinant) clade and codon positions 9, 15, 16, 24, 30 and 138 for clade O. The SLAC and FEL methods detected a fewer number of codon sites under positive selection for clade O and no codon sites for the N (and recombinant) clade (Table 1). None of the three methods detected codon sites under positive selection for clade C. In addition, clade N and the recombinant clade were analysed separately for the whole CP coding region, using both PAML and HyPhy. These supplementary analyses did not provide additional candidate codon sites submitted to positive selection. On the contrary, the previously-identified codon sites were detected with lower statistical support, which was probably due to the fact that splitting the sequence dataset in two subsets decreased the statistical power of analyses. The fact that different amino acid positions in the CP, depending on the clades, appeared to be under positive selection suggests that selective pressures and/or substitution rates at those particular positions could vary between clades. To confirm this possibility, a LRT (Knudsen and Miyamoto 2001) was used to detect evolution rate shifts between the different PVY clades at the codon positions previously detected as candidates for positive selection with both PAML and HyPhy. Codon position 25 was shown to evolve significantly more rapidly in clades C, N or in the recombinant clade than in clade O (Table 2). Also, codon position 26 evolved faster in clades O or N than in clade C, while codon position 68 evolved faster in the recombinant clade than in clades C or O. Finally, codon position 10 evolved faster in clade O than in clade C. No significant rate shift was observed between clades for codon position 11, which was detected as positively selected in clade O (Table 1). Overall, these results add support to most of the codon positions candidate for positive selection and indicate that positive selection at individual codon positions is specific for one or several PVY clades. Application of Knudsen and Miyamoto’s (2001) LRT to the other nucleotide positions of the CP coding region did not reveal more positions showing evolutionary rate shifts than under random (at the 5% type-I error threshold).To validate the prediction that the previously-identified codon positions are evolving under positive selection, we measured the effect of nonsynonymous substitutions at these positions on the virus fitness. For this purpose, PVY clones differing only by these mutations should be compared. No cDNA infectious clones corresponding to PVY isolates from clade O or from the recombinant clade are available. Consequently, we used the cDNA clone from isolate N605 (Jakab et al. 1997; Bukovinszki et al. 2007), belonging to clade N, and we selected the codon positions detected as candidates for positive selection in clade N and the recombinant clade both by HyPhy and PAML, i.e. codon positions 25 and 68 (Table 1). At position 25, 55 isolates of 62 from clade N or the recombinant clade, including Version définitive du manuscrit publié dans / Final version of the manuscript published in : Molecular Biology and Evolution, 2011, 28 (9), 2707-2717 DOI: 10.1093/molbev/msr105. The publisher version is available at http://mbe.oxfordjournals.org/content/28/9/2707.full.pdf+html. 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 N605, possessed an asparagine (Asn) codon, four isolates possessed an isoleucine (Ile) codon, while three isolates possessed a serine or a lysine codon. At position 68, 50 isolates of 62, including N605, possessed a glutamic acid codon (Glu), eleven isolates possessed a lysine (Lys) codon, while one isolate possessed a valine codon. Only the effects of the most frequent amino acid substitution observed in clade N and in the recombinant clade at these two positions were examined, i.e. effects of the Asn25Ile and Glu68Lys codon substitutions. Effect of mutations at positively-selected codon positions on PVY fitness in different host plants We compared the relative accumulation of N605 and its mutants N605-CP25 and N605-CP68, which differ from N605 only by the Asn25Ile or by the Glu68Lys codon substitutions in the CP coding region, in tobacco and potato, the two major cultivated hosts of PVY isolates from clade N (Kerlan and Moury 2008; Moury 2010). Two different inoculation protocols were used. In competition experiments, N605 and either N605-CP25 or N605-CP68, were co-inoculated manually to the plants using inocula composed of similar concentrations of the PVY variants. In single inoculation experiments each virus variant was inoculated manually to a separate set of plants after calibration of the concentration of the virus solutions. In these latter experimental conditions, virus accumulations were compared by DAS-ELISA (Janzac et al. 2010) and it was therefore important to examine the influence of the Asn25Ile and Glu68Lys substitutions in the CP of N605 on the reactivity of the different PVY variants toward the polyclonal antibodies used in DAS-ELISA. Using purified virus of known absolute concentration we did not observe any significant differences of DAS-ELISA reactivity among the three PVY variants (Kruskal-Wallis test, p-value=0.35). N605-CP25 and N605-CP68 showed average DAS-ELISA reactivities 14% and 18% lower than N605, respectively. However, four independent virus purification experiments could be insufficient to detect a true difference in DAS-ELISA reactivity between PVY variants. Consequently, the ELISA data were analysed both with and without correction by the relative reactivity factor between variants. In the Xanthi cultivar of tobacco, the N605-CP25 mutant accumulated at a significantly higher rate than N605, both in competition situation (p-value=0.008; Fisher exact test) or after single inoculation (Fig. 2). In this host, there was no accumulation difference between N605 and the N605-CP68 mutant whatever the inoculation procedure (with or without competition between the two variants) (Fig. 2). In the Bintje cultivar of potato and in the single inoculation procedure, N605-CP68 accumulated at a much higher rate than N605 (about 10 times more), while there was no accumulation difference between N605 and N605-CP25 (Fig. 2). In this host, results from competition experiments differed from results obtained after single inoculation: N605 was more competitive than either the N605CP25 or the N605-CP68 mutant (p-value=0.04 and p-value<0.0001, respectively; Fisher exact tests; Fig. 2). Importantly, the relative competitiveness of each mutant against N605 was significantly different in the two host plants (p-value<0.0001 whatever the plant species; Fisher exact tests). In addition, there was no significant difference of infectivity in Xanthi plants among the three PVY variants N605, N605-CP25 and N605-CP68, when measuring the proportion of infected plants after manual inoculation with serially-diluted solutions of calibrated virus concentration (data not