Durability of plant major resistance genes to pathogens depends on the genetic background, experimental evidence and consequences for breeding strategies

Authors


Author for correspondence:
Alain Palloix
Tel:+33 43272 2741
Email: alain.palloix@avignon.inra.fr

Summary

  • • The breakdown of plant resistance by pathogen populations is a limit to the genetic control of crop disease. Polygenic resistance is postulated as a durable alternative to defeated major resistance genes. Here, we tested this postulate in the pepper–Potato virus Y interaction.
  • • The virus was selected for virulence towards monogenic and polygenic host resistance, using serial inoculations in laboratory and in natural epidemic conditions. The frequency of resistance breakdown and the genetic changes in the virus avirulence gene were analysed.
  • • The monogenic resistance provided by the pvr23 gene was defeated at high frequency when introgressed in a susceptible genetic background whereas it was not when combined to partial resistance quantitative trait loci. The suppression of emergence of virulent mutants because of the genetic background resulted both from a differential selection effect and the necessity for the virus to generate multiple mutations. The virus adaptation to the polygenic resistance required a step-by-step selection with a primary selection for virulence towards the major gene, followed by selection for adaptation to the genetic background.
  • • Polygenic resistance proved more durable than monogenic resistance, but breeding strategies giving priority to major resistance factors may jeopardize the progress in durability expected from polygenic resistance.

Introduction

The breakdown of genetic resistance by plant pathogen populations is a major limit to the genetic control of crop disease. Increased research efforts have been made to propose breeding strategies, resistance gene or cultivar deployment strategies and cultivation methods that aim at controlling the pathogen evolution over time and its adaptation to the resistant cultivars (Kiyosawa, 1982; Finckh et al., 2000; Lindhout, 2002; Pink, 2002). The evolutionary potential of the pathogen population, when inferred from its population genetic and dynamic characteristics, aids prediction of the risk of pathogen evolution towards virulence and of breakdown of resistance or other control methods (McDonald & Linde, 2002; Garcia-Arenal & McDonald, 2003). However, for a given pathogen, this does not predict if one resistant host genotype will be more rapidly overcome than another. Plant breeders who face a pathogen population have to optimize the plant genotype by choosing the most promising resistance genes and gene combinations for durability of resistance.

Very few experimental data presently explain why, in a given pathosystem, some resistant cultivars succeed over a long time while others rapidly fail. The fitness penalty associated with virulence acquisition is strongly suspected to be involved in the durability of resistance since it determines the increase in frequency of virulent genotypes and their survival in the presence as well as the absence of the resistant cultivars (Leach et al., 2001). This was demonstrated experimentally by Vera Cruz et al. (2000) in the Xanthomonas oryzae–rice pathosystem. However, fitness penalty can also be transitory since compensatory mutations may increase the fitness of virulent genotypes in the long term (Garcia-Arenal et al., 2001; Wijngaarden et al., 2005). The number of mutations required for virulence acquisition was another parameter that was related to resistance durability by Harrison (2002). He compared the relative durability of resistance in several plant–virus pathosystems and showed that the greater the number of mutations required for virulence, the more durable was the resistance. Indeed, the probability of emergence of virulent mutants is directly dependent on the probability of occurrence of these mutations, which may affect resistance durability. Recent experimental data supported the importance of these parameters in the pepper–Potato virus Y (PVY) interaction. The observed durability of the resistance alleles at the pvr2 locus was related to the number of mutations required for virulence in the virus (Ayme et al., 2007). In addition, for a given pvr2 allele, the emergence of different virulent PVY mutants was proportional to the relative fitness and to the relative rates of the virulence mutations (Ayme et al., 2006). Results from several authors consolidate the idea that population genetics and dynamics may deliver new criteria for breeding durable resistance. The most durable resistance genes are those that require multiple mutations from the pathogen for virulence, with mutations causing the highest fitness penalty.

In addition to these formal analyses, many empirical results suggest greater durability of polygenic resistance relative to monogenic resistance (Parlevliet, 2002). Highly durable polygenic resistance was observed in several pathosystems where monogenic resistance failed (Turkensteen, 1993; Chen et al., 2003; Schurnbusch et al., 2004). However, a few analyses showed that pathogen populations may adapt to polygenic resistances, bypassing all or part of the resistance quantitative trait loci (QTL) (Le Guen et al., 2007) [Correction added on 14 May 2009, after first online publication: the citation year for ‘Le Guen et al.’ was changed from ‘2006’ to ‘2007’]. In this case, resistance bypassing was observed a posteriori (i.e. the isolated pathogen populations were assumed to result from the earlier deployment of resistant cultivars) but the dynamics of their evolution could not be observed. To date, no experiment clearly demonstrates the higher durability of polygenic resistance, or permits one to address the question of the determinism of this durability. Two main hypotheses are currently proposed:

  • • the greater the number of resistance factors to breakdown, the greater the number of virulence mutations required in the pathogen genome and the less probable their occurrence;
  • • the selection pressure owing to quantitative resistance factors is lower than that of major genes and does not permit the emergence of virulent mutants from the pathogen population.

In this paper, we addressed this question by comparing the relative durability of a monogenic and a polygenic resistance in controlled conditions, in a pathosystem that permits analysis of the genetic changes of the pathogen during its adaptation to the resistance. In the pepper–PVY interaction, both the avirulence gene of the virus and the corresponding resistance gene in the plant were isolated, and the point mutations controlling the specificity of the pathotype–cultivar interaction were identified (Ruffel et al., 2002; Moury et al., 2004). One resistance allele, pvr23, displayed a weak durability and the breakdown frequency of this allele was high when introgressed in a susceptible genetic background but null when this allele was combined to partially resistant genetic background carrying resistance QTL. The additive effect of these QTL was weak, since it only delayed systemic infection and symptom expression. However, the genetic background displayed a major effect on the emergence of pvr23-virulent mutants. The selection of virulent variants remained possible through sequential selection (i.e. first by the major resistance factor and further by the combination of the major resistance factor with the resistance QTL) resulting in the progressive accumulation of the virulence mutations. These results suggested rules for the durable management of resistance genes and QTL.

Materials and Methods

Pepper genotypes used in this work were Capsicum annuum L. inbred lines with differential resistances to PVY isolates: Yolo Wonder (allele pvr2+) is susceptible to all isolates, Florida VR2 (allele pvr22) is resistant to PVY pathotype (0,1,3) and Perennial carries a polygenic resistance, including the allele pvr23, that confers resistance to pathotype (0,1,2) and resistance QTL conferring partial resistance to the other isolates. The pvr2 locus was mapped on the P4 chromosome, and three main QTLs with resistant alleles originating from the resistant accession Perennial were mapped on the chromosomes P1 (two QTL) and P6, using a PVY isolate derived from the SON41 strain that overcame the pvr23 resistance allele (Fig. 1; Caranta et al., 1997). These three QTL displayed individual effects (R2) > 10%. Three doubled haploid lines issued from the F1 hybrid (Perennial × Yolo Wonder) were used and selected on the basis of molecular markers at the pvr2 locus (Rubio et al., 2008) and at markers flanking the three main resistance QTL (Fig. 1). The three selected pepper lines were: HD285 carrying the allele pvr23 with susceptible alleles at the three QTL (susceptible genetic background), designated here as Rs; HD233 carrying the resistant allele pvr23 with resistant alleles at the three QTL (resistant genetic background), designated here as Rr; and HD223 carrying the susceptible allele pvr2+ with resistant alleles at the three QTL (resistant genetic background), designated here as Sr.

Figure 1.

Genetic mapping of the pvr2 locus and of the quantitative trait loci (QTL) for resistance to PVY in pepper and graphical genotypes of the pepper inbred lines used in the present work. The pvr2 locus and QTL are located according to the QTL analysis of Caranta et al. (1997) and the pepper genetic map of Lefebvre et al. (2002). The black and areas colours on chromosomes P4, P1 and P6 represent, respectively, the Perennial and Yolo Wonder alleles. pvr2+ and pvr23 are, respectively, the susceptible and resistant alleles; s and r are used to indicate the susceptible versus resistant alleles at QTL.

The PVY isolates and infectious clones were derived from the SON41p isolate collected in France in 1982. The SON41p isolate belongs to pathotype (0,1,2) and is virulent towards Yolo Wonder and Florida VR2, but not towards Perennial, Rs or Rr. Here, we define virulence as the genetic ability of a pathogen to cause a compatible interaction with a host genotype leading to disease. Five SON41p mutants were isolated from Rs and gained virulence towards the pvr23 allele as the result of single amino acid substitutions in the central part of the Viral Protein genome linked (VPg) (Ayme et al., 2006). These mutants were named S101G, T115K, T115R, D119N and S120C according to the mutation in the VPg that characterized them. Infectious cDNA clones from the SON41p isolate (Genbank accession AJ439544) and from the VPg mutants were obtained by Moury et al. (2004) and Ayme et al. (2006).

Artificial inoculation tests were carried out in climate-controlled rooms. Viruses were propagated in Nicotiana spp. to obtain high-titre inocula for tests of C. annuum. Leaf tissue from infected Nicotiana spp. plants developing severe symptoms of the disease was homogenized in four volumes of 0.03 m phosphate buffer (pH 7.0) supplemented with 2% (w : v) diethyldithiocarbamate, 20 mg ml−1 of active charcoal and 20 mg ml−1 of carborundum. Test plants with two expanded cotyledons (2–3 wk after sowing) were inoculated manually on their cotyledons. Evaluation of virus infection was performed by double-antibody sandwich enzyme-linked immunosorbent assay (DAS-ELISA), as described by Legnani et al. (1995) at various time-points after inoculation. Symptom severity was evaluated using a semiquantitative scale placing each individual plant into three classes: 1, no visible symptom; 2, weak symptoms; 3, severe mosaic or necrotic symptoms. Symptom intensity was checked every week for 5 wk after inoculation and the area under the symptom progress curve (AUSPC) was calculated with the formula:

image

(Si is the symptom intensity at the date ti, in days). Only the DAS-ELISA positive plants were considered for the AUSPC value calculation. For the assessment of the pathogenicity of PVY isolates and infectious clones, the same procedure was used, except that the relative concentration of the inoculum extract was previously determined by semiquantitative DAS-ELISA as in Ayme et al. (2006) and the extracts were adjusted to the same concentration using dilution in the extraction buffer before inoculation. For each pepper genotype–PVY isolate combination, 50 plantlets were inoculated and individually assessed for symptom and DAS-ELISA. Three to four weeks after inoculation, relative virus concentration of the test plants was determined by semiquantitative DAS-ELISA.

For serial inoculation experiments, successive sowings of Perennial and Rs were performed at 5-wk intervals. Fifty plantlets per genotype were artificially inoculated as previously described by the SON41p cDNA clone and the five PVY mutants (S101G, T115K, T115R, D119N and S120C). Successive passages were achieved by extracting the inoculum from systemically infected leaves pooled from 10 plants per genotype, and inoculating this extract to the 50 plantlets of the same genotype for the next passage. At each passage, the plants were individually tested by DAS-ELISA to assess the infection percentage 4 wk after inoculation. After four successive passages in Perennial and in Rs the virus was extracted from two pools of 20 plants per genotype and PVY mutant. These virus populations were compared with their initial cDNA clones for the percentage infection and severity of symptoms in Yolo Wonder, Rs and Perennial (50 inoculated plantlets per virus population). The viral RNA was extracted from the same virus extracts and the sequence of the VPg cistron was established as in Moury et al. (2004).

Controlled inoculations of PVY in cultivation conditions were performed in two independent insect-proof tunnels. In each tunnel, 100 plants of Rr (possessing pvr23 and known resistance alleles at QTL) were planted to evaluate the ability of the virus to adapt to the polygenic resistance. Six control plants were added: two Florida VR2 (susceptible to SON41p but not to the D119N mutant), two Rs (resistant to SON41p, but susceptible to D119N) and two Yolo Wonder (susceptible to both PVY variants). The inoculum source was introduced in each tunnel by planting two Florida VR2 plants infected by SON41p and two Rs plants infected by the PVY mutant D119N. This pvr23-virulent mutant was chosen because of its prevalence (53%) over the four other mutants in experimental conditions, and it was also frequently detected in open fields (Ayme et al., 2006). Two weeks after plantation, aphids (Myzus persicae) were deposited on the PVY-inoculated plants, in order to favour the transmission of the virus to healthy plants. Symptom observations and ELISA test of every plant were performed at 15-d intervals from June to October, during the 18 wk covering the cultivation season. When infected plants were detected, viral RNA was extracted and the sequence of the VPg cistron was established to determine which isolate was present and if additional mutations were detected as described in Moury et al. (2004).

Results

Relative breakdown frequency of monogenic versus polygenic resistance

In order to compare monogenic and polygenic resistances for their relative frequency of breakdown, young plantlets of the doubled haploid line Rs (possessing the pvr23 resistant allele in a susceptible genetic background), of Perennial (possessing the pvr23 resistant allele in a partially resistant background including the resistance alleles at the three QTL) and of the susceptible Yolo Wonder were inoculated with the SON41p infectious clone of PVY that is avirulent towards pvr23. Fifteen independent tests including 20–50 plants of each genotype were performed; in total 332 Yolo Wonder, 332 Rs and 471 Perennial plants were checked for systemic PVY infection by DAS-ELISA (Table 1). All the Yolo Wonder plants displayed strong mosaic symptoms at 2–3 wk after inoculation and were ELISA positive, whereas only 76 Rs plants were ELISA positive and displayed systemic symptoms (mosaic and necrosis) 4–5 wk after inoculation. None of the 471 inoculated Perennial plants were positive in ELISA or displaying any symptoms. When virus extracts from infected Rs plants were used as inocula, all back-inoculated Rs plants were infected 10–14 d post inoculation, indicating a virulence gain of the virus population towards the pvr23 allele. A total of 29 of these virus extracts were analysed in previous studies (Ayme et al., 2006, 2007) and the analysis showed that the virulence gain towards the pvr23 allele resulted from single amino acid substitutions that occurred at different positions in the VPg of the virus genome (namely S101C T115K, T115R, D119N and S120G) that conferred virulence towards the pvr23 allele.

Table 1.  Frequency of infection as attested by DAS-ELISA tests in three pepper genotypes: Yolo Wonder (PVY susceptible), Rs (pvr23 allele in a susceptible genetic background) and Perennial (pvr23 allele in a partially resistant genetic background), 5 wk after inoculation with a pvr23-avirulent Potato virus Y (PVY) infectious clone (SON41p)
Cultivar (and resistant genotype)Yolo Wonder (pvr2+)Rs (pvr23)Perennial (pvr23 + resistant genetic background)
  • 1

    Data from 15 independent tests, each including 20–50 plants of every genotype. The frequency of Rs infected plants differed from that of Perennial and Yolo Wonder plants at P < 10−6 (Fisher's exact test).

  • 2

    Confidence intervals cannot be calculated when observed standard deviations are zero.

Number of inoculated plants1332332471
Number of infected plants1332 76  0
Mean % infection (± SD)100 (± 0) 27.6 (± 9.3)  0 (± 0)
95% confidence interval)2 23.3–31.9

Selection of PVY isolates for virulence towards the polygenic resistance through serial inoculations

The five cDNA clones possessing the sequence of the SON41p clone and one of the five substitutions conferring virulence towards pvr23 were tested for their ability to adapt to the polygenic resistance of Perennial. These PVY clones were submitted separately to repeated passages on Perennial plants and were maintained in parallel through repeated passages in Rs plantlets. The initial cDNA clones uniformly infected 100% of the Rs plants over the serial inoculations. Figure 2 shows that three of the five mutants (T115K, T115R and D119N) infected 60–100% of Perennial plants systemically as early as the first passage. For these three mutants, the infectivity remained stable over the repeated passages. The other two mutants were unable to infect Perennial at the first passage, so the S120C inoculum source had to be re-extracted from the Rs plants for the second passage. For the same reason, the S101G inoculum for the third passage was also extracted from the Rs plants. The T115K, T115R and D119N mutants were submitted to four effective passages in Perennial, whereas S120C and S101G were submitted to three and two effective passages in Perennial, respectively.

Figure 2.

Time-course of infection of five PVY variants across repeated passages in pepper genotype Perennial. For the S120C and S101G variants, the inoculum source had to be extracted from Rs plants (HD285) at the second and third passages, respectively, since no Perennial plants were infected in earlier passages.

After the passages, each PVY population from Perennial and from Rs was compared with its original cDNA clone for severity of symptoms in Yolo Wonder, Rs and Perennial hosts, for the frequency of infection of Perennial plants and for their VPg sequence (Table 2). No differences in symptom severity were observed between the original infectious clones and their derived isolates (data not shown) with severe symptoms in Yolo Wonder plants (susceptible) as well as in Rs plants, but late and weak symptoms in Perennial. Table 2 shows that the infection rate in Perennial was also not significantly affected by the repeated passages for the mutants T115R, T115K or D119N, but that it increased significantly (P < 0.05) for the mutants S120C and S101G. Sequence data from the VPg of the isolates after repeated passages in Rs or in Perennial showed no differences with the initial PVY clone for T115K or T115R. An additional mutation T115M was gained in the D119N, S120C and S101G clones after repeated passages in Rs. In these last two clones the additional mutation D119N was specifically gained only after repeated passages in Perennial.

Table 2.  Amino acid differences in the central part of the VPg of Potato virus Y (PVY) isolates and frequency of infection of Perennial plantlets after serial inoculation on distinct pepper hosts
Original infectious cloneHost genotype used for serial inoculationsAmino acid sequence of the central part of VPg% Infection of Perennial*
  1. *For each PVY clone, frequencies followed by different letters are significantly different at P = 0.05 (Fisher's exact test). VPg, Viral Protein genome linked.

SON41PSEVRRKHVEDDEIETQALDSH  0
S101GG--------------------  0 a
RsG-------------M------  0 a
PerennialG-------------M---N-- 38 b
T115K--------------K------100 a
Rs--------------K------ 86 a
Perennial--------------K------ 96 a
T115R--------------R------100 a
Rs--------------R------ 84 a
Perennial--------------R------100 a
D119N------------------N-- 72 a
Rs--------------M---N-- 76 a
Perennial------------------N-- 78 a
S120C-------------------C-  0 a
Rs--------------M----C-  0 a
Perennial------------------NC- 60 b

Selection of PVY isolates for virulence towards the polygenic resistance in glasshouse production conditions

The ability of PVY SON41p and of the pvr23 virulent mutant D119N to break the polygenic resistance down in epidemic conditions (i.e. aphid transmission of the virus to adult plants) was assessed in plastic tunnels. Both PVY clones were introduced into the insect-proof tunnels planted with the Rr pepper line possessing the allele pvr23 and partial resistance QTL from Perennial. Aphid vectors (Myzus persicae) were introduced into the tunnels, in order to transmit the primary inoculum to the healthy plants (secondary infections) as happens in cultivation conditions. The epidemics progressed in the tunnels with the growth of the aphid population and the Yolo Wonder plants (fully susceptible) first displayed secondary infections with either SON41p or D119N, attesting that both isolates were transmitted to healthy plants. At 8–12 wk, the SON41p isolate was also detected in most Florida VR2 plants and the D119N isolate in most of the Rs plants. Virus infection in Rr plants was detected only 12 wk after aphid release, as attested by ELISA. At the end of the trial (18 wk), only four Rr plants (two plants in each tunnel) displayed PVY infections that were associated with severe mosaic symptoms that progressed into plant necrosis. Sequence data of the VPg from the four isolates collected in these plants showed that they all possessed the D119N mutation in their VPgs and the additional mutation T115K. This indicated that they probably derived from the D119N mutant and not from the SON41p isolate. The additional mutation T115K occurred independently in the two tunnels suggesting that it was related to the adaptation of PVY mutant D119N to the polygenic resistance of Rr.

One of these four isolates was used for a second year trial and selection in the same conditions. It was artificially inoculated to two Rr plantlets that were further planted together with healthy Rr plants in the insect-proof tunnels. Twelve weeks after the aphid release on the inoculated plants, the aphid population spread all over the plots and secondary infections were observed in Yolo Wonder and Rs control plants, but also in a few Rr plants. At the end of the experiment, 75% and 87%, respectively, of the Rr plants were infected, as indicated by an ELISA test in each tunnel, and showing severe systemic symptoms. Isolates were collected from 20 of these plants (10 in each tunnel) and their VPg cistron was sequenced. All the isolates displayed the same sequence with both D119N and T115K mutations. No other mutations were detected within the VPg cistron compared with the initial SON41p infectious clone.

Pathogenicity of the PVY isolate selected in the insect-proof tunnels

One isolate from the second year experiment (isolate 53) was compared with the initial infectious clone possessing the single mutation D119N and with a clone possessing the two mutations D119N and T115K for its pathogenicity towards a host range in the laboratory (Figs 3, 4). For each pepper genotype–PVY isolate/mutant combination, 100% of the 50 plantlets were infected 3 wk after inoculation, except for the Perennial–D119N combination (82% infection after 4 wk). Considering AUSPC values, symptom severity was high for all the PVY isolates in Yolo Wonder (fully PVY susceptible) as well as Rs (allele pvr23 in susceptible genetic background). In Sr (susceptible allele pvr2+ but resistance alleles at the QTL) the single and double mutants from cDNA clones induced weak and late symptoms whereas isolate 53 displayed a higher AUSPC value. In Rr and Perennial (pvr23 and resistant alleles at the QTL) the AUSPC values were significantly different between the single mutant (weak and late symptoms), the double mutant (intermediate AUSPC values) and isolate 53, which displayed earlier and more severe symptoms. The relative virus concentrations at 3 wk after inoculation were similar in the susceptible pepper Yolo Wonder whatever the virus, but higher for the double mutant and/or isolate 53 compared with the single mutant in the other resistant genotypes. Isolate 53 displayed a significantly higher concentration than the double mutant in Rs and particularly in Perennial.

Figure 3.

Comparison of mutants and isolate derived from the SON41p PVY clone for pathogenicity on pepper genotypes in artificial inoculation conditions. Vertical bars of histograms are standard errors. For each pepper genotype, PVY strains with different letters display significantly different area under the symptom progress curve (AUSPC) values at P < 0.05 (Tukey HSD test). Open bars, D119N; hatched bars, D119N + T115K; closed bars, isol 53.

Figure 4.

Virus titre as estimated from enzyme-linked immunosorbent assay (ELISA) readings of sap dilution curves from different pepper genotypes inoculated with PVY clones and isolates derived from SON41p 3-wk after inoculation. For each dilution level, absorbance values with different letters display are significantly different at P < 0.05 (Tukey HSD test). Circles, D119N; squares, D119N + T115K; triangles, isol 53.

Discussion

Our experimental data demonstrated that the frequency of breakdown of a resistance governed by a major gene was high when introgressed in a susceptible genetic background, whereas this breakdown did not occur when the same gene was introgressed in a partially resistant genetic background. Previous analyses (Ayme et al., 2006, 2007) showed that every time the Rs plants were systemically infected after SON41p inoculation, it corresponded with a resistance-breaking event caused by one single amino acid substitutions in the central part of the VPg cistron (29 events analysed). In these Rs systemic infection events , back-inoculations of Rs plants also showed that the PVY isolate gained virulence, and we assume that the infection of 76 Rs plants over the 332 inoculated was caused by the breakdown of the resistance governed by pvr23 by an adapted viral variant. The Rs line differed from Perennial by the alleles at the known resistance QTL, but also at many other loci in the genetic background. Both the QTL detected in Caranta et al. (1997) and the remaining genetic background may be responsible for the suppression of emergence of the virulent mutants. Thus, this effect will be globally assigned to the genetic background (including known QTL) surrounding the major gene. Such results may be extended to numerous plant resistances, since most of the polygenic resistances to diseases were shown to result from the combined effect of one major resistance QTL controlling 50–70% of the resistance and several minor resistance QTL (Lefebvre & Chèvre, 1995; Young, 1996). The higher durability of polygenic resistances is commonly hypothesized and has been observed a posteriori in a few cases (Lindhout, 2002; Parlevliet, 2002). This is the first experimental evidence derived from a direct comparison in controlled conditions. How the genetic background or the resistance QTL control the emergence of virulent variants towards the major resistance factor can be explored.

Two main hypotheses regarding the higher durability of polygenic resistance can be addressed here. The durability may result from the requirement of multiple mutations to break down the multiple resistance factors, in which case it is related to the initial step of virus evolution and strictly depends on the probability of occurrence of multiple mutations. An alternative hypothesis is that, after their appearance, selection of virulent variants is more efficient or rapid in plants with monogenic resistance than polygenic resistance. This relates to the further steps of virus evolution (i.e. differential selection (competition) between virulent and avirulent variants during plant infection). Note that these two hypotheses are not mutually exclusive.

Under artificial inoculation of pvr23-virulent PVY clones onto young plantlets, three of these clones (T115K, T115R and D119N) proved able to infect the Perennial plantlets carrying polygenic resistance. When SON41p was inoculated to Rs plants (pvr23 in a susceptible background) these variants were selected at frequencies from 2.5% (for T115R) to 20% (for D119N) (Ayme et al., 2006) but they never emerged in the 471 Perennial plants inoculated in the same conditions. Together, these results indicate that these variants were not selected in Perennial although they were present at low frequency in the initial virus population. This strongly argues in favour of the differential selection hypothesis for increased durability of pvr23 as a result of additional factors in the host genetic background.

A different pattern of evolution was observed in epidemic conditions (i.e. adult pepper plants and aphid virus transmission). In such cultivation conditions, only 2% of the polygenically resistant peppers were infected, which was lower than in artificial inoculation of young plantlets (78% in Table 2). Moreover, isolates from the four infected Rr plants exhibited the T115K mutation in addition to the initial D119N mutation. This additional mutation occurred independently in the two tunnels, suggesting that the double mutation conferred an advantage for virulence towards the polygenic resistance in field conditions. Indeed, when inoculated to the pepper host range, the PVY clone with the double mutation T115K + D119N displayed higher AUSPC values and virus concentrations in Rr and Perennial plants compared with the single D119N mutant. In the second-year experiment, the tunnel isolate further proved able to infect 75–87% of Rr plants, confirming its gain in virulence towards polygenic resistance. Moreover, the resulting isolate (53) displayed a higher virus concentration than the PVY clone with the double mutation when inoculated to Rs and Perennial and higher AUSPC values in Rs, Sr, Rr and Perennial plants. Note that Perennial showed higher resistance than Rr plants (Fig. 3), which indicates that resistance QTL additional to the three already mentioned but undetected in the analysis of Caranta et al. (1997) are present in the genetic background of Perennial. Whatever the number of QTL, the higher virulence of isolate 53 in plants possessing the major gene and/or resistance QTL, compared with the PVY clone with the double mutation in the VPg, clearly indicates that additional mutation(s) in other genome regions of this isolate confer a higher virulence towards the major gene, the resistance QTL from the genetic background and their combination. The direct observation of adaptation of virus populations to the resistant host revealed the progressive changes that lead to the gain of virulence, as was observed a posteriori by Le Guen et al. (2007) [Correction added on 14 May 2009, after first online publication: the citation year for ‘Le Guen et al.’ was changed from ‘2006’ to ‘2007’] who isolated Microcyclus ulei strains that bypassed enlarged sets of QTL after large deployment of resistant rubber trees. Together, these results show that the gain of virulence towards the polygenic resistance in field conditions requires multiple mutations, and argues in favour of our first hypothesis that at least one additional mutation within the VPg but also additional mutation(s) in other genome parts were required for virulence towards polygenic resistance in fields. The combination of the D119N and T115K amino acid substitutions in the VPg plus additional mutations in other genome regions is unlikely to occur in a few host passages where virus replication is very low because of resistance. This explained why the adaptation to the polygenic resistance was not achieved directly from the SON41p initial population but required a step-by-step selection for virulence, first towards the major gene and further towards the combination of the major gene and QTL.

Despite the higher durability of the polygenic resistance compared with the monogenic one, the ability of the virus to respond to a progressive selection has to be considered when breeding resistant cultivars. Breeding for resistance in most crops often consists in introgressing major genes or major QTL from a donor exotic germplasm into recipient elite genotypes that are susceptible to the disease. This introgression is processed through successive backcrosses by the recipient genotype, and molecular marker-assisted backcrosses (MAB) accelerate the process and optimize recovery of the recipient genetic background (Michelmore, 1995; Hospital & Charcosset, 1997). For genes such as pvr23 this strategy would facilitate the breakdown of resistance by providing an evolutionary springboard for further adaptation of the virus population to more complex resistance, jeopardizing any further genetic progress expected from combinations with resistance QTL. A step-by-step evolution has been found when pyramiding major resistance genes that were individually defeated in previous cultivars, leading to multivirulent pathogen genotypes (Pink, 2002). Resistance to PVY in pepper germplasm is very frequent with 35.6% of resistant accessions among the 884 tested (Sage-Palloix et al., 2007) and a large diversity of pvr2 alleles (Charron et al., 2008). The pvr23 allele was identified in several locally cultivated populations from Asia (Perennial) and Central America (CM334); in these genotypes the allele was combined with other resistance genes or QTL (Dogimont et al., 1996; Caranta et al., 1997). These local open-pollinated cultivars were selected and maintained by farmers who operated mass selection in fields for yield, quality and low symptom expression. This mass field selection in regions where potyviruses are prevalent probably led to the combination of complementary genetic factors for resistance efficiency and durability. Breeding for durable resistance in modern cultivars should promote the selection for major resistance factors together with appropriate genetic background. Further analysis of the genetic factors impeding the emergence of pvr23-virulent PVY populations and their relationship with resistance QTL will deliver criteria to select for such genetic backgrounds.

Acknowledgements

We are grateful to the Comité Technique Permanent de la Sélection (CTPS, Ministry of agriculture) and to the Clause Vegetable seeds, Gautier, Rijk Zwaan, Vilmorin and Sakata seed companies for financial support and technical participation to the program. We also thank Ghislaine Nemouchi, Pascale Mistral and Vincent Simon for their excellent technical assistance, and Nathalie Boissot for critical review of the manuscript. The experiments were conducted in compliance with the current laws of France for bio-security and confinement of GMOs.

Ancillary