Comparison to the population-based study

I compare here the results of the species wide study to the results of the population-based study and to previous population-based studies of the same genes in the same set of species. Most signatures of selection, which were observed at the population level (purifying selection at the reference dataset, balancing selection at Pto and Pfi), were confirmed by the species wide study in S. peruvianum and could even be extended to S. chilense (or

S. corneliomulleri). This may imply that selective pressures causing the observed signatures of purifying selection at reference loci are homogenous throughout the species, while acting locally within each population. This is not surprising considering that the reference loci are genes with known or likely housekeeping function (Baudry et al. 2001; Roselius et al. 2005;

Städler et al. 2005). I suggest that it is improbable that selective constraints on these

housekeeping functions differ substantially between populations of the same species and therefore the observed signature of selection within single populations should reflect the overall selection pattern within the whole species. Note however the difference in selective constraint observed between species (these results and Tellier et al. 2011).

Interestingly, at the R genes studied, the population-based study and the species wide

sample yielded similar results, pointing to balancing selection as major force driving evolution at Pto and Pfi. These findings potentially indicate that the selective constraint

resulting in the observed pattern is homogenous throughout the whole species or at least throughout the sampled range. However, pathogen incidence as well as parameters of

coevolutionary dynamics such as parasite prevalence and disease severity are in general not a homogenous pressure, but usually harbour spatial and temporal variation within and among populations (Laine & Tellier 2008; Thompson 2005). This heterogeneity together with spatial structuring of host and parasite populations can cause direct frequency-dependent selection and as a consequence the pattern of balancing selection observed at coevolving genes in the entire metapopulation (Brown & Tellier 2011; Gavrilets & Michalakis 2008). Population structure is a key factor in promoting coevolutionary dynamics leaving the signature of balancing selection in the entire species (Burdon & Thrall 1999; Thrall & Burdon 2002). If coevolutionary dynamics occurring due to direct frequency-dependent selection are present for long periods of time, the genomic signature of balancing selection would be seen, even if local plant populations may coadapt to the local pathogenic genotypes. In this case, allele composition and allele frequencies would likely differ between populations.

Alternatively, it is also possible that the two sampling schemes do not capture the heterogeneity in pathogen pressure between populations. It is possible that in some populations and parts of the species range, pathogen incidence is high (hot spot of coevolution) but it is low or pathogens are absent in other part of the range (cold spots, Thompson 2005). If this heterogeneity subsists for a long enough period of time, the signature of balancing selection will be seen both in sequences at hot spots (i.e. Tarapaca population) and at the species wide level. Such possible geographic variation in pathogen pressure resulting in differential outcomes of host-pathogen coevolution has been suggested for example for the Cf-2R gene in S. pimpinellifolium (Caicedo 2008). Note that as a corollary, balancing selection might not be detected both in samples at cold spots and the species wide level, depending on gene flow in the metapopulation.

Finally, a third plausible explanation for the observed genomic signatures could involve a selective constraint promoting polymorphism patterns resembling that of balancing selection, because it is not imposed by a pathogen but rather some cellular function of the two genes. This, however, is unlikely at least in the case of Pto for which a potential for host-

pathogen coevolution has been demonstrated functionally (Bernal et al. 2005; Rose et al.

2005; Rose et al. 2007).

In the case of the Rin4 gene, the two sampling schemes yielded different results, but

these are not necessarily contradictory. A pattern of strong purifying selection was present in the species wide sample, while the pattern observed in the population-based study could have been due to either purifying or positive selection. I hypothesize that the Rin4 gene is in

beneficial mutations may occur occasionally and may sweep through the local population. If gene-flow between demes is low, this mutation might not immediately migrate to other demes and can as a consequence not immediately increase in frequency in the entire species (Charlesworth et al. 1997; Whitlock 2003). Alternatively, this new mutation may not confer a

fitness advantage in other local environments and therefore be counter-selected in other populations. Both scenarios could lead to the pattern observed in the two studies: Rin4 seems

to be conserved in the entire species, but might experience positive selection in a local population. Of course, it is also possible that Rin4 is conserved throughout the entire species

and the pattern observed in the Tarapaca population is simply due to genomic peculiarities around this gene, such as low recombination rate, and/or close relationship between individuals within populations combined with low statistical power to detect outlier genes.

One great advantage of the species wide sample approach is that it allows investigating the evolutionary history of the three resistance genes beyond species boundaries. Analyzing the patterns of sequence variation between closely related species helps to understand details of their evolutionary history and to extend the evolutionary time scale under investigation. The population-based study revealed signatures of balancing selection at the Pto and Pfi genes as did the species wide sample. However, only the sampling of the collecting phase of the metapopulation coalescent tree in both S. peruvianum and S. chilense