Adaptation to the most abundant host genotype in an agricultural plant–pathogen system – potato late blight



    1. INRA, Agrocampus Rennes, Université Rennes, UMR1099 BiO3P (Biology of Organisms and Populations Applied to Plant Protection), F-35653 Le Rheu, France
    Search for more papers by this author
  • I. GLAIS,

    1. INRA, Agrocampus Rennes, Université Rennes, UMR1099 BiO3P (Biology of Organisms and Populations Applied to Plant Protection), F-35653 Le Rheu, France
    Search for more papers by this author

    1. INRA, Agrocampus Rennes, Université Rennes, UMR1099 BiO3P (Biology of Organisms and Populations Applied to Plant Protection), F-35653 Le Rheu, France
    Search for more papers by this author

    1. INRA, Agrocampus Rennes, Université Rennes, UMR1099 BiO3P (Biology of Organisms and Populations Applied to Plant Protection), F-35653 Le Rheu, France
    Search for more papers by this author

D. Andrivon, INRA, Agrocampus Rennes, UMR1099 BiO3P (Biology of Organisms and Populations applied to Plant Protection), F-35653 Le Rheu, France.
Tel.: (+33) 223 485193; fax: (+33) 223 485180; e-mail:


This study investigated local adaptation of Phytophthora infestans populations, the causal agent of potato late blight, to two susceptible potato cultivars, each grown for a number of years and over large areas in separate French regions. We measured aggressiveness (quantitative pathogenicity) of each pathogen population to sympatric and allopatric hosts in a reciprocal cross-inoculation experiment. There was no evidence for specific host adaptation in this pathosystem. At both local and regional scales, the distribution of aggressiveness fits a pattern of adaptation to the most common host genotype. Our observations support the theoretical predictions that large pathogen dispersal rates and genetic drift, revealed by the comparisons of the genotypic structures of the populations tested, can lead to a local adaptation pattern detectable only at a large spatial scale. The unravelling of adaptive patterns at different spatial scales can be used for a more efficient management of the disease.


Two main types of adaptive patterns can occur in populations of parasites unable to actively choose their hosts: adaptation to the most common host genotype, corresponding to a single specialist phenotype optimally adapted to one host and less fit on other hosts, or specific local adaptation (or maladaptation) to different host genotypes, corresponding to a set of specialist phenotypes each maximizing fitness on one host and leading to a highly fragmented pathogen population (Gandon et al., 1996, 1998; Kawecki & Ebert, 2004). Both cases actually correspond to adaptation to locally prevalent hosts, but acting at different spatial scales. While specific local adaptation reflects divergent selection on separate host genotypes with little gene flow between pathogen populations, adaptation to the most common host can be observed whenever a single host genotype is largely prevalent within the dispersion range of the pathogen. Consequently, whether the pathogen is locally adapted (resp. maladapted) or displays adaptation to the most common host depends primarily on the balance between response to selection, fitness costs to adaptation to different hosts, gene flow and genetic drift (Gandon & Van Zandt, 1998; Kaltz & Shykoff, 1998). Because parasites have larger population sizes, shorter generation times and higher rates of mutation and migration than their hosts, they are usually expected to be locally adapted to sympatric hosts (Hamilton et al., 1990; Ebert & Hamilton, 1996; Lively, 1996; Kaltz & Shykoff, 1998; Gandon & Michalakis, 2002). Several experiments, in which parasites performed better on sympatric than on allopatric hosts, support this assumption; however, such evidence is not always detected, and several data sets even point to local maladaptation of the parasite (see for reviews Kaltz & Shykoff, 1998; Kawecki & Ebert, 2004; Greischar & Koskella, 2007).

Testing if parasite populations display local (mal)adaptation should ideally be done by comparing adapted parasite populations to their presumably less adapted ancestors. Because this is usually not possible, adaptive patterns are most often tested via reciprocal cross-infection experiments, and more generally by reciprocal-transplant experiments (Lively, 1996; Kaltz & Shykoff, 1998). There are two distinct ways to test for local adaptation in a reciprocal-transplant experiment (Gandon & Van Zandt, 1998): (i) by comparing the performance of a parasite population in different environments/hosts (the ‘home vs. away’ criterion of Kawecki & Ebert, 2004) or (ii) by comparing the performance of a parasite population in its native environment with that of other parasite populations transplanted there from different environments (the ‘local vs. foreign’ criterion of Kawecki & Ebert, 2004). In both cases, the relevant statistical criterion is the significance of the interaction term between pathogen origin and host origin: local (mal)adaptation results in a significant interaction (Greischar & Koskella, 2007).

Evolutionary forces shaping pathogen population structures is an important research topic in evolutionary ecology in natural ecosystems, but also in agro-ecosystems; indeed, besides answering key evolutionary questions, testing local adaptation in host–pathogen systems can provide new or improved strategies for the control of diseases through host genetic resistance (Mc Donald & Linde, 2002). Several characteristics of agricultural pathosystems make them well suited for the analysis of host–parasite adaptation. First, cultivated hosts are frequently made of genetically identical (or homogeneous) plants, such as F1 progenies of homozygous parents or vegetatively propagated clones, allowing one to test the prediction that parasites should be more locally adapted on asexual than on sexual hosts (Ladle, 1992). Just as importantly, host populations are often grown as single cultivars deployed over large agricultural areas, which make them highly vulnerable to rapidly evolving parasites (Brown, 1994). The spatial distribution of agricultural cultivars is therefore usually a monoculture of a single genotype at the plot or regional scale, and a mosaic of monocultures at the nation or continental scale: this allows for the investigation of adaptive processes at spatial scales ranging from fields to regions to countries. Finally, because large monocultures of short-living hosts might represent the same temporal stability (and thus persistence of selection) as a single long-living host, parasite adaptation to asexual hosts can be conceptually interpreted in the same way as local adaptation of parasites to long-living, sexually reproducing hosts (Gandon et al., 1998).

We took advantage of these characteristics to investigate local adaptation using the potato late blight pathosystem involving the oomycete parasite Phytophthora infestans and its annual host plant Solanum tuberosum. Late blight is one of the most destructive diseases of the cultivated potato, and occurs wherever potato is grown (Harrison, 1992). Potato cultivars are vegetatively propagated clones, and each potato production area is usually dominated for several years or decades by one host cultivar, different between regions. As a consequence, potato crops at a regional scale are akin to persistent monocultures. On the pathogen side, P. infestans has a short generation time (3–5 days) and a large asexual multiplication rate (Harrison, 1992), resulting in polycyclic epidemics which favour rapid response to selection. In France, one of the two mating types required for sexual reproduction has been rare until very recently, so populations were primarily collections of closely related clonal lineages (Lebreton et al., 1998; Montarry et al., 2006). Simulation and analytical studies suggest that local adaptation depends on the relative migration rates of host and pathogen: parasites which migrate more than their hosts tend to be locally adapted (Gandon et al., 1996; Gandon, 2002). Because the dispersion capacity of P. infestans, via airborne asexual sporangia containing the infective zoospores, is larger than the dispersion capacity of its sessile host S. tuberosum (Bourke, 1964), local adaptation is expected. Finally, the narrow host range of P. infestans in temperate countries should also favour local adaptation in this host–pathogen agro-ecosystem (Lajeunesse & Forbes, 2002).

Previous work showed that P. infestans populations collected from susceptible and partially resistant cultivars grown side by side at single sites do not show specific local adaptation (Montarry et al., 2006), but that genetically isolated populations show a clear adaptation to locally prevalent cultivars (Andrivon et al., 2007). In the present work, we wanted to assess whether genetic isolation or prolonged selection was the main driver of local adaptation. This can be investigated by designing cross-inoculation experiments between P. infestans populations collected from different production areas, each dominated by a single cultivar for several decades and separated by several hundreds of kilometres, making population migration from one region to the other within a given growing season possible but unlikely to be massive. In this study, we used such a design and measured aggressiveness (i.e. quantitative pathogenicity –Vanderplank, 1963; equivalent to the term ‘virulence’ as used by evolutionists and ecologists –Jarosz & Davelos, 1995; Poulin & Combes, 1999) of two pathogen population on sympatric and allopatric host populations. Here, we consider as the sympatric host the cultivar from which any given isolate was collected and as allopatric hosts any cultivar included in the test and different from the cultivar of origin of the isolate.

Materials and methods

Phytophthora infestans isolates

To investigate adaptive patterns of P. infestans to sympatric and allopatric cultivars of its host at a regional spatial scale, a total of 133 single-lesion isolates was sampled in 2004 and 2005 from two production areas. Field populations P-O and P-B were sampled in western France and northern France, respectively. Population P-O consisted of isolates collected in a circular area of approximately 10 km diameter, centred on the GPS point 48°22′N–4°45′W (west of Brest) and dominated for many years by the susceptible cultivar Ostara. Population P-B included isolates collected in a circular area of approximately 80 km diameter, centred on the GPS point 50°24′N–2°47′E (between Amiens and Lille), and dominated by the susceptible cultivar Bintje. P-O isolates (24 isolates from five fields and 45 isolates from 10 fields in 2004 and 2005 respectively) were all collected from cv. Ostara, whereas P-B isolates (23 isolates from seven fields and 41 isolates from nine fields in 2004 and 2005 respectively) were from cv. Bintje. The geographical centres of the sampling regions are distant by approximately 600 km.

In order to test adaptive patterns also at a local spatial scale, a second collection was established from naturally infected leaflets of both Ostara and Bintje plants present in three experimental plots placed in the two production areas described above, and at the INRA potato breeding station at Ploudaniel (48°30′N–4°19′W), where there is no dominant cultivar. In both years and in each location, the trials contained two replicate plots (two rows of six plants each) of each cultivar. One hundred and nineteen single-lesion isolates were sampled from these experimental plots during the 2 years, on the basis of approximately 10 isolates by cultivar, location and year.

Infected leaf tissue was collected each year during the first phase of the epidemic. Single-lesion isolates were established and maintained as axenic cultures on pea agar as previously described (Montarry et al., 2006).

Mating type determination

Since P. infestans is a heterothallic organism, isolates from opposite mating types (A1 and A2) need to be present simultaneously on the same plant/organ for the pathogen to undergo sexual reproduction. We determined the mating type of each isolate by pairing on pea agar with known A1 and A2 testers, incubating in the dark at 18 °C for 10–14 days, and observing cultures for oospore formation under a microscope (Shaw et al., 1985). Isolates forming oospores with the A1 tester were rated as A2 mating type and those that formed oospores with the A2 tester were rated as A1 mating type.

Pathogenicity tests

Potato plant material

Bintje and Ostara plants were grown from certified seed tubers in 13-cm pots (one tuber per pot) filled with 1 : 1 : 1 sand–peat–compost mixture, in a glasshouse regulated at 15–20 °C (night/day temperatures) and 16 h of photoperiod. Potato cultivars being clones propagated vegetatively, these seed tubers were genetically identical to those planted in the plots or fields sampled. Plants were watered with a nutrient solution (Hakaphos; NPK 15/10/15) once a week.

Inoculum preparation

Maintaining isolates as axenic cultures on agar media for extended periods can alter their aggressiveness; however, pathogenicity can be restored by re-infecting leaf tissue (Jeffrey et al., 1962; Jinks & Grindle, 1963). Each isolate was thus multiplied separately on its own cultivar of origin to produce the inoculum used for aggressiveness determination. To this end, detached leaflets of potato cultivars Ostara and Bintje were collected from plants grown in the greenhouse for 6–8 weeks from healthy, certified seed tubers, and deposited on the lids of inverted Petri dishes containing 12 g L−1 water agar, acting as humid chambers. They were infected by depositing a suspension of P. infestans sporangia collected by flooding a 3-week-old culture with 10 mL sterile tap water and gently scraping the colony surface to remove sporangia. Before inoculation, sporangial suspensions were kept at 4 °C for approximately 4 h to promote zoospore liberation. After 10 days of incubation in controlled conditions (18 °C/15 °C day/night temperatures, 16 h daylight), the sporangia produced on infected leaflets were collected in sterile water, and the concentration of the resulting suspensions were adjusted to 5 × 104 sporangia per mL using a haemocytometer.

Aggressiveness quantification

Each isolate was tested for aggressiveness on each host cultivar (Ostara and Bintje). Six leaflets were detached from 6/8-week-old plants and placed abaxial face up on the lids of inverted Petri dishes containing 12 g L−1 water agar (two leaflets per dish). Each leaflet was infected by depositing a 20 μL drop of the sporangial suspension as close as possible to the leaflet centre. Inoculated leaflets were then incubated for 7 days as described above. The latency period (LP) was determined by observing daily the apparition of sporangia. Seven days after inoculation, lesion size was measured in two perpendicular directions and lesion area (LA) was calculated assuming an elliptic shape. Then, leaflets were washed in 10 mL Isoton II (saline buffer; Beckman, Coulter France, Villepinte, France), and the sporangia production (SP) per leaflet was determined with a Coulter Z2 counter (Beckman Coulter France, Villepinte, France). A composite aggressiveness index (Ai) was calculated for each isolate and leaflet using the formula Ai = ln (LA × SP × 1/LP). This index is indicative of the epidemiological potential of an isolate. Similar aggressiveness indices have been used for this and other pathogens (Crute et al., 1987; Thakur & Shelly, 1993; Day & Shattock, 1997; Flier & Turkensteen, 1999).

Molecular characterization

P-B and P-O isolates were grown in pea broth autoclaved for 20 min at 120 °C. After 10–15 days of incubation at 18 °C, mycelium was washed three times in sterile water, and lyophilized. DNA was extracted as described by Lebreton et al. (1998) and stored in TE buffer containing 10 m Tris–HCl and 0.1 m EDTA (pH 8.0). DNA concentration and purity were estimated using a spectrofluorimeter (SpectraMax M2; Packard BioScience, Meriden, CT, USA).

Ten polymorphic microsatellite loci: Pi4B, Pi4G and PiG11 developed by Knapova & Gisi (2002); and Pi02, Pi04, Pi16, Pi33, Pi56, Pi63 and Pi70 recently developed by Lees et al. (2006), were amplified for allele detection. Microsatellite Polymerase Chain Reactions (PCR) were performed in a 12.5 μL volume containing between 20 and 200 ng of DNA of P. infestans, 2.5 μL of 5× PCR Buffer (Promega France, Charbonnières les Bains, France), 0.3 mm of each dNTP, 2.5 mm of MgCl2 (Promega), 0.3 μm each of forward and reverse primers, and 1.25 U of Taq DNA polymerase (GoTaq® flexi DNA polymerase, Promega). PCR was performed in a MJ Research thermocycler under the following conditions: the PCR started with a cycle of 2 min at 95 °C, followed by 30 cycles of 20 s at 95 °C, 25 s at 56 °C (for PiG11), 58 °C (for Pi02, Pi04, Pi16, Pi33, Pi56, Pi63, Pi70 and Pi4B) or 60 °C (for Pi4G) and 60 s at 72 °C, and finished with an elongation cycle of 5 min at 72 °C. In order to detect simultaneously the alleles at several loci, primers were labelled with three fluorescent dyes: FAM (PiG11, Pi33, Pi 63, Pi70, Pi02 and Pi4B), NED (Pi56, Pi04, Pi4G) and HEX (Pi16). Amplification products were pooled into three groups, based on expected allele sizes: PiG11, Pi56 and Pi33; Pi63, Pi04 and Pi70; and Pi02, Pi4G, Pi16 and Pi4B respectively. Ten μL samples, comprising 9.84 μL of deionized formamide Hi-DiTM (Applied Biosystems, Courtaboeuf, France), 0.06 μL of 400 HD ROXTM Size standard (Applied Biosystems), and 0.1 μL of PCR multiplexed product, were loaded into an ABI Prism 3130xl DNA sequencer run according to manufacturer’s instructions (Applied Biosystems). DNA fragments were automatically sized with the GeneMappertm 3.5 software (Applied Biosystems).

Data analyses

Molecular data

A number of isolates showed three or four alleles at one or more of the 10 microsatellite loci, confirming the extensive variation in ploidy in European P. infestans populations (Tooley & Therrien, 1987). As allele detection was not quantitative, it was impossible to adequately estimate allele frequencies. The microsatellite data were thus used as multilocus fingerprints, reflecting the presence or absence of each allele at each locus for each isolate. The resulting binary matrix was subjected to a cluster analysis with the unweighted pair-group method with arithmetic averages (upgma), with Dice’s coefficient as a metric (Nei & Li, 1979), using the BioNumerics software (Applied Maths BVBA, Kortrijk, Belgium). Cophenetic correlations were calculated for each node to estimate the stability of clusters.

Diversity was estimated for each population using the Shannon index (Sheldon, 1969), calculated as:


where pj is the frequency of the jth genotype in the population. Diversities between populations were statistically tested using a two-tailed t-test (Hutcheson, 1970).

Direct pairwise comparisons of populations were performed using the Rogers index (Groth & Roelfs, 1987), calculated as:


where pj1 and pj2 represent the frequencies of genotype j in population 1 and in population 2, respectively, for all genotypes present in population 1 and(or) population 2. Hr varies from 0, when all genotypes are present at equal frequencies in the two populations, to 1, when the two populations do not share any single genotype. Intermediate values of Hr arise either from very dissimilar frequencies of the same genotype in the two populations, from the presence of some different genotypes between the populations, or from a combination of both (Andrivon, 1994).

Aggressiveness data

The effects of pathogen population (production area of sampling), test cultivar and their interaction on aggressiveness indexes relative to P-B and P-O isolates were tested for each year through an anova, using the GLM procedure in the statistical analysis system software (SAS Institute Inc., Cary, NC, USA). Contrast tests allowed ‘home vs. away’ (for each pathogen population) and ‘local vs. foreign’ (on each test cultivar) tests.

To differentiate between adaptation to production areas and to cultivars, aggressiveness data from isolates sampled in experimental field plots were analysed using a second anova model. The effects of pathogen population (location of sampling), cultivar of origin, and test cultivar, as well as their two-way and three-way interactions, on aggressiveness indexes were tested separately for each year.

A third two-way anova model was used to explore the relationship between aggressiveness and genotypic structure by testing cluster effect, host population effect and the corresponding two-way interaction.


Mating type

All 153 isolates from western France (70 in 2004 and 83 in 2005), including those from P-O and from the experimental plots placed in the production area Ostara and in Ploudaniel, belonged to the A1 mating type. Conversely, the 99 isolates from northern France (39 in 2004 and 60 in 2005), including those from P-B and from the experimental plots placed in the production area Bintje, contained a significant proportion of A2 isolates (20.5% in 2004 and 16.7% in 2005). Despite similar overall frequencies in both years, A2 isolates were found in only two of the seven fields sampled in 2004, but were recovered from all nine fields sampled in 2005.

Genotypic structure

Genotypic diversity

Fifty-eight genotypes were found among the 133 isolates of P. infestans tested. The P-B population (Hs = 3.337) was significantly more diverse than the P-O population (Hs = 2.497; t121 = 4.03, P < 0.01). Pairwise comparisons of populations using the Rogers index revealed important variations in genotype frequencies between years in a given production area (HR2004/2005 = 0.767 for the P-O population, and HR2004/2005 = 0.870 for the P-B population), as well as between production areas in a given year (HRP-O/P-B = 0.833 for 2004, and HRP-O/P-B = 0.911 for 2005).


Forty-five different alleles were identified over the 10 microsatellite loci. The phenetic analysis of the multilocus fingerprints revealed six clusters (Fig. 1). Interestingly, cluster V was composed exclusively of A2 mating type isolates, although four A2 isolates belonged to other clusters (two A2 isolates in cluster I and one each in clusters III and IV). The largest genotypic distance between isolates, based on Dice’s coefficient, was 64.15%. All P-O isolates were present in only three different clusters (II, III and IV). By contrast, P-B isolates were distributed across all six clusters. Some genotypes were detected in only one of the 2 years, suggesting that the populations were subject to an extinction/recolonization dynamics during the winter (elimination of several genotypes during the inter-crop phase and migration from neighbouring sources in early spring), as previously observed in other parts of Europe (Zwankhuizen et al., 2000).

Figure 1.

 Clustering of Phytophthora infestans genotypes in populations P-B and P-O from two consecutive years (2004 and 2005). Genotypic distances between clusters were estimated using an upgma algorithm applied to the matrix of Dice similarity index, based on allele distributions at 10 polymorphic microsatellite loci. Figures at each node of the tree are cophenetic correlations.


Regional spatial scale

The absence of interaction between pathogen populations and test cultivars for each year (Table 1) revealed no evidence for local adaptation in this pathosystem at the regional spatial scale investigated (between the two pathogen populations, P-B and P-O, and the two host cultivars, Bintje and Ostara). Both populations were more aggressive on Bintje than on Ostara in the four comparisons (two sites in 2 years), although the contrast tests for aggressiveness differences on their own (home) vs. other host populations (away) were significant in 2005 only (Table 1 and Fig. 2). Population P-B was more aggressive than population P-O on both Bintje and Ostara in three out of the four comparisons, although contrast tests for aggressiveness differences of resident (local) vs. nonresident pathogen populations (foreign) were significant in 2004 only (Table 1 and Fig. 2).

Table 1.   Regional spatial scale: analyses of variance for aggressiveness index in each year, 2004 and 2005, using the GLM procedure in the sas software.
d.f.Mean squareF-valueP > Fd.f.Mean squareF-valueP > F
  1. Sources of variation are the pathogen population effect, the test cultivar effect and the corresponding two-way interaction. Contrasts show the Home vs. Away comparisons (H. vs. A.) and the Local vs. Foreign comparisons (L. vs. F.).

  2. The statistically significant effects are indicated as *P < 0.05 and ***P < 0.001.

Main effects
 Pathogen population114.833061.50< 0.0001***10.07770.330.5681
 Test cultivar11.26965.260.0241*14.602119.39< 0.0001***
Interaction effects
 Pathogen pop. * test cv 10.31121.290.259010.52722.220.1380
 Error900.2412  1680.2374  
 H. vs. A. (pathogen pop P-O)11.44994.020.050811.05615.240.0245*
 H. vs. A (pathogen pop P-B)10.15851.360.249913.939214.230.0003***
 L. vs. F. (test cv. Bintje)15.423533.44< 0.0001***10.50483.000.0869
 L. vs. F. (test cv. Ostara)19.720630.36< 0.0001***10.10010.330.5692
Figure 2.

 ‘Home vs. away’ and ‘local vs. foreign’ comparisons at the regional spatial scale. Graphs show, for each year, the aggressiveness index of the two Phytophthora infestans populations (P-B and P-O) on each test cultivar (means ± SE). Statistically significant differences are indicated as *P < 0.05 and ***P < 0.001.

Local spatial scale

The absence of two-way (cultivar of origin * test cultivar), and three-way (cultivar of origin * test cultivar* pathogen populations) interactions in each year (Table 2) showed no evidence for local adaptation at the local spatial scale. Wherever the pathogen population came from and whatever the cultivar of origin, isolates collected in all experimental plots were more aggressive on Bintje than on Ostara in both years, the ‘home vs. away’ contrasts being significant in three of the four comparisons (Table 2 and Fig. 3). In both years, isolates collected from Bintje in the experimental plots were more aggressive than isolates collected on Ostara, wherever the pathogen population came from and whatever the cultivar on which it was tested, but only one of the ‘local vs. foreign’ contrasts was significant (in 2005 on test cultivar Ostara; Table 2 and Fig. 3). Pathogen population and test cultivar effects were exactly concordant with those observed in the populations from agricultural fields at the regional scale (Tables 1 and 2).

Table 2.   Local spatial scale: analyses of variance for aggressiveness index in each year, 2004 and 2005, for isolates from experimental field plots in each location, using the GLM procedure in the sas software.
d.f.Mean squareF-valueP > Fd.f.Mean squareF-valueP > F
  1. Sources of variation are the cultivar of origin effect, the pathogen population effect, the test cultivar effect and all corresponding interactions. Contrasts show the Home vs. Away comparisons (H. vs. A.) and the Local vs. Foreign comparisons (L. vs. F.).

  2. The statistically significant effects are indicated as *P < 0.05, **P < 0.01 and ***P < 0.001.

Main effects
 Cultivar of origin (1)11.78313.220.075611.90588.850.0037**
 Pathogen population (2)23.42046.170.0029**20.03170.150.8631
 Test cultivar (3)13.04125.490.0209*16.678431.02< .0001***
Interaction effects
 (1) * (2)21.45652.630.076720.19140.890.4142
 (1) * (3)10.01430.030.872610.19660.910.3415
 (2) * (3)20.01020.180.832620.40271.870.1593
 (l) * (2) * (3)20.18100.330.722220.00590.030.9732
 Error1120.5544  1020.2153  
 H. vs. A. (cv. origin Ostara)11.84802.100.151814.637122.46< .0001***
 H. vs. A. (cv. origin Bintje)11.58986.020.0174*12.12829.870.0027**
 L. vs. F. (test cv. Bintje)10.80611.880.175610.44553.070.0852
 L. vs. F. (test cv. Ostara)10.75560.970.327611.76986.390.0144*
Figure 3.

 ‘Home vs. away’ and ‘local vs. foreign’ comparisons at the local spatial scale. Graphs show, for each year, the aggressiveness index of Phytophthora infestans isolates sampled in the experimental field plots on the two potato cultivars (cv. origin Bintje and cv. origin Ostara) on each test cultivar (means ± SE). Statistically significant differences are indicated as *P < 0.05, **P < 0.01 and ***P < 0.001.

Relationship between aggressiveness and genotypic structure

Isolates from cluster V were significantly less aggressive than those in other clusters, accounting for the significant cluster effect (F5,254 = 3.62, P < 0.01) in the anova. All isolate clusters were more aggressive on the host population Bintje than on the host population Ostara (F1,254 = 11.59, P < 0.01), explaining the absence of a significant interaction between clusters and host populations (F5,254 = 0.52, P = 0.76). Although isolates from cluster V (composed only of A2 mating type isolates) were less aggressive than those from other clusters, no aggressiveness differences were detected between A1 and A2 isolates (data not shown).


The main objective of this paper was to test adaptive patterns of a polycyclic parasite to sympatric and allopatric cultivars of its host. The absence of statistical interactions between pathogen populations and test cultivars at the regional scale and between cultivar of origin and test cultivar at the local scale shows that there is no specific local adaptation (or maladaptation) in this pathosystem at the spatial scales studied here. This result confirms and extends previous observations from local field plots (Montarry et al., 2006), despite theoretical expectations (i.e. high pathogen migration rate, larger pathogen population size and shorter pathogen generation time –Hamilton et al., 1990; Ebert & Hamilton, 1996; Lively, 1996; Kaltz & Shykoff, 1998; Gandon & Michalakis, 2002). At both the local and regional scales, the adaptive pattern emerging from the data is consistent with the hypothesis of adaptation to the host genotype Bintje, corresponding to populations mainly composed of a single specialist pathogen phenotype, optimally adapted to one host cultivar and less fit on others. This conclusion accounts for the observations relative to both field and experimental plot populations, which showed (i) an overall higher aggressiveness of P. infestans populations on Bintje than on Ostara and (ii) an overall higher aggressiveness of P. infestans populations collected on Bintje than those collected on Ostara.

Local adaptation is most likely to occur when differences in size and quality of habitats are small (Kisdi, 2002). We do not think that susceptibility differences in Bintje and Ostara alone could explain the lack of local adaptation reported here. Indeed, despite the higher susceptibility of Bintje relative to Ostara in controlled conditions optimal for the development of the pathogen, no significant difference in the disease progress curves were detected between these two cultivars based on weekly assessments of the mean late blight severity of all plants in each sampled plot for six to eight consecutive weeks (data not shown). We therefore favour the hypothesis that the adaptation to cultivar Bintje observed in this and the previous study (Montarry et al., 2006) results from the much more widespread distribution of Bintje, relative to Ostara. Indeed, the production area dominated by Bintje (approximately 20 000 km2;) is much larger than the one dominated by Ostara (approximately 300 km2;). Moreover, Bintje can be found all over the national territory, in agricultural fields or in private gardens, whereas Ostara is more strictly restricted to its small production area.

The lack of specific host adaptation at local and regional scales, observed here, is most likely also the consequence of extensive gene flow between populations across regions, coupled with the wider distribution of Bintje. This hypothesis is consistent with (i) the distribution of the genotypes from the P-B population into all phenetic clusters, explaining the lack of a strong structure in French P. infestans populations and confirming previous reports (Lebreton et al., 1998) and (ii) the high migration capacity of this pathogen, which was postulated from the first European epidemic of late blight in 1845 (Bourke, 1964). The high genotype diversity observed in the P-B population could be the consequence of sexual reproduction, A1 and A2 isolates being present simultaneously in this population. Alternatively, genotype diversity could also be the effect of the P-B geographical situation, at the confluence of flows from other large potato production areas (UK, the Netherlands, Belgium).

Adaptive patterns provide an insight into the spatial scale at which (co)evolutionary dynamics occur. A number of studies report no difference between parasite performance on sympatric and allopatric hosts (Parker, 1989; Ennos & McConnell, 1995; Davelos et al., 1996; Dufva, 1996), possibly because local adaptation may be present at a larger or a smaller scale than the investigation considers. The host–pathogen association studied here confirms hypotheses that large pathogen dispersal rates and genetic drift, which together generate strong extinction/colonization dynamics within local populations (Zwankhuizen et al., 2000), can lead to a local adaptation pattern detectable only at a large spatial scale (Andrivon et al., 2007). Indeed, previous studies (Montarry et al., 2006) have shown that at the field plot scale, P. infestans populations from a susceptible potato cultivar (Bintje) are more aggressive than those from a partially resistant cultivar (Désirée), with no differential adaptation to either host. Flier et al. (2007) reported similar results between European P. infestans populations from the moderately resistant cv. Santé and the susceptible Bintje. Those results, similar to the pattern observed here, can be due to either adaptation to Bintje because of its susceptibility (i.e. the selection of the most aggressive isolates is faster on Bintje than on other cultivars, because the pathogen population size is larger on a susceptible cultivar), or because of its overall abundance in the host population. The local adaptation pattern observed between P. infestans populations from France and Morocco (Andrivon et al., 2007), i.e. in geographically isolated populations with no or very limited gene flow, highlighted that local adaptation was mainly due to the abundance of potato cultivars, the Moroccan population being locally adapted to a partially resistant cultivar. This is consistent with P. infestans life history traits (i.e. long distance migration of asexual spores, strong genetic drift during inter-epidemic phases and response to selection) which should prevent local adaptation in locally or regionally connected populations but favour them between genetically isolated populations (Kaltz & Shykoff, 1998).

Although both criteria (‘home vs. away’ and ‘local vs. foreign’) used to detect local adaptation are not independent, they are also not equivalent regarding the underlying mechanisms they reveal. The first detects a trade-off in performance across habitats, and thus gives evidence of adaptive costs, whereas the second shows differences in selection pressure across habitats. Therefore, variation in host resistance can obscure tests using the ‘home vs. away’ comparison, whereas variation in pathogen aggressiveness can bias the ‘local vs. foreign’ comparison, as shown by Thrall et al. (2002) in the Linum marginaleMelampsora lini interaction. Moreover, in order to predict the outcome of the co-evolutionary conflict between host and parasite or to interpret patterns of local adaptation, it is important to have information on the population structure, usually obtained from neutral molecular markers, of the two partners (Delmotte et al., 1999). Here, six of the eight ‘home vs. away’ comparisons are statistically significant, whereas only three of the eight ‘local vs. foreign’ comparisons are. This suggests that adaptive costs are likely to occur in this pathosystem. Such costs can lead to local adaptation in genetically separated populations of P. infestans (Andrivon et al., 2007), but are possibly too low to structure populations on different hosts if gene flow is important, as seems to be the case here.

The absence of a specific local adaptation pattern does not necessarily mean that the process of divergent selection is not operating (Kawecki & Ebert, 2004). Metapopulation dynamics may strongly interfere with local selection process and temporal variation in selection prevents local adaptation (Kassen, 2002). Therefore, detecting local adaptation in metapopulations, where colonization and extinction play significant roles (Hanski, 1999), is problematic (Gandon & Van Zandt, 1998; Kawecki & Ebert, 2004). The large extinction rates during winter in P. infestans (Shattock, 1976; Zwankhuizen et al., 2000) and the variation in local frequencies of pathogen genotypes over successive seasons observed here and in previous work (Lebreton et al., 1998) point to a significant effect of genetic drift in shaping local population structures, and thus prevent local adaptation to be detected at that scale (Thrall & Burdon, 1997; Burdon & Thrall, 2000). Large parasite dispersal rates decrease the degree of adaptation of the parasite to local hosts by introducing locally maladapted genotypes into each parasite population; as a consequence, if migration completely swamps local dynamics, local adaptation may only be apparent at higher geographical scales (Gandon & Van Zandt, 1998; Kaltz & Shykoff, 1998).

The unravelling of adaptive patterns at each of these spatial scales can now be put to use for a more efficient management of the disease by appropriate deployment strategies of potato cultivars at local and regional scales. Such patterns should be based on minimizing cultivar homogeneity and pathogen population flows. The erratic performance of cultivar associations observed in the past (Garrett et al., 2001; Andrivon et al., 2003) and their strong dependence on disease pressure (Pilet et al., 2006) is consistent with gene flow as a key factor in adaptive patterns in this pathogen. Therefore, preserving the stability of resistance in time and space in this pathosystem will require both the diversification of resistance sources across the potato sole and adequate management of inoculum sources, through sanitation of seed, rotation and/or elimination of refuse piles (large piles of cut or diseased tubers, unsuitable for storage or marketing and left over in the vicinity of fields where they serve as major sources of primary inoculum; see for instance Zwankhuizen et al. (2000) for a quantitative assessment of their role as sources of primary foci in subsequent crops). The development of demogenetic models is now required to best assess the evolutionary and demographic consequences of management options, such as habitat fragmentation by growing multiple cultivars or by favouring a more rapid turn-over of these cultivars in time, and to best combine them according to agroclimatic constraints.


We thank L. Dubois, R. Pellé and J.M. Abiven for establishing field plots and help with sampling; E. Lesné for establishing isolate collections, H. Douchy and B. Marquer for technical assistance in the biological tests, and G. Mialdéa for help with the molecular characterization. Genotyping of the isolates was performed using the Ouest-Genopole® facilities. We also thank O. Kaltz and three anonymous referees for very helpful comments on earlier versions of this paper. J.M. was supported by a PhD grant from Bretagne-Plants, acting on behalf of the ACVNPT (Association des Créateurs de Variétés Nouvelles de Pomme de Terre); support from the European Integrated Project BioEXPLOIT [Contract 513959 (FOOD)] is also gratefully acknowledged.