This is the first time that a detailed population genetic analysis of P. infestans in the northern Andean region was conducted. Until now, the pathogen population structure was unknown in this portion of the continent (Colombia and Venezuela). Colombia is situated in a critical geographical position, acting as a bridge between Central and South America. According to FAO statistics, this country has a commercial history in trading potatoes with several countries in Europe, Asia, North and South America, and the Caribbean . This suggests there have been opportunities for migration of the pathogen, along with its host, towards and within the northern Andean region.
The low genetic diversity found on global and regional scales in the Northern Andean region (Colombia and Venezuela), in both mitochondrial and nuclear regions, is consistent with previous reports in other countries of South America . Even for microsatellite markers low levels of genetic diversity have been reported in Colombia and Venezuela [25–27]. Specifically, for Ras, only 2 out of 13 haplotypes found globally could be observed in the region. The A2 mating type was discovered in Ecuador and Colombia [16, 26], and sexual reproduction is expected to produce new allele combinations. However, even though the A1 and A2 mating types converge in the same geographical region, as occurred in Cundinamarca Department (Colombia), sexual recombination is apparently not prevalent. This is probably due to a process of host adaptation [11, 16, 48], or because the presence of the A2 mating type is too recent to have resulted in recombination. Additionally, the low frequency of the A2 mating type described in Colombia diminishes the chances for sexual reproduction to occur . Indeed, in the nuclear genes few individuals were heterozygous, suggesting that genetic interchange occurs but at extremely low frequency. Because of this, the presence of heterozygous individuals might be the result of ancestral polymorphisms or gene flow and not the result of sexual reproduction.
The gene networks for the different regions showed, as a general pattern, that the relationships between the haplotypes in each population of the Northern Andean region are consistent with variation within a clonal population and that haplotypes are discriminated by one or two steps. However, in the Southern Andes, at least for β-tubulin and Avr3a, more than two steps separated four and two haplotypes, respectively. The more divergent haplotypes were found in samples from S. betaceum and corresponded to the mitochondrial haplotype Ia. Recently it has been suggested that a new species, P. andina, with an Ia mitochondrial haplotype, could be associated with S. betaceum[8, 23, 49]. However, its taxonomic status has not been clarified yet, and continues to be considered as P. infestans sensu lato. The presence of this new species in the population of the southwestern Andes could explain the different pattern observed in Avr3a and β-tubulin.
The genetic variation of P. infestans is mainly explained by the variability present within each population that is generated by the existence of different low-frequency alleles, particularly for β-tubulin and Avr3a in the Southern Andes. Nevertheless, the variation detected in this study was enough to show some genetic differentiation. At the regional level, the P. infestans population in Venezuela appeared to be isolated from the Central and Eastern Andes populations. Indeed, at the mitochondrial level Venezuelan isolates belonged largely (11 out of 12) to the Ia haplotype, and just one corresponded to IIa, the most common haplotype in the North Andean region. This and the presence of a particular haplotypic composition for the nuclear genes may suggest that the Venezuelan population probably has been structured by different population events. Reasons for the apparent genetic isolation of the Venezuelan isolates of P. infestans have been discussed elsewhere . The estimates of genetic flow support the idea that the population of Venezuela is not donating or receiving migrants from any other region (Table 5). However, a large area has not been sampled on the Colombian side close to the Venezuelan border. More sampling is therefore needed in this region in order to detect possible recent gene flow.
Additionally, two significant patterns were observed. First, subdivision was found between the Eastern Andes with the rest of Colombia for β-tubulin. This could be the result of a long history of self-sufficiency in terms of potato seed tuber supply in the eastern Andes as well as avoiding movement of the pathogen on plant tissue. Second, results obtained with the mitochondrial gene Cox1 and nuclear genes suggested that historically the Southwestern and the Eastern P. infestans populations have been isolated, but recent gene flow could be taking place.
The selection imposed on a gene with a known function in host recognition as Avr3a may have resulted in a different pattern of sequence polymorphisms in the sampled population in comparison to the other nuclear regions analyzed. However, low genetic diversity was found for this gene. No genetic subdivision could be detected in the Northern Andean region, in contrast with what was found for β-tubulin and Ras. At the amino acid level interesting patterns emerged. Previous studies reported only three polymorphic positions in a sampling of isolates from S. tuberosum from different locations worldwide. Two alleles were characterized based on those amino acid positions, C19 K80 I103 and S19 E80 M103, with avirulent and virulent phenotypes, respectively . Here we report four new allelic variants. Three were only present in the isolates from the southwestern Andes. The natural selection analysis at the amino acid level showed that three out of the nine polymorphic positions were under diversifying selection and two of these were located in the C-terminus region as previously reported . According to the model of interaction of the virulence genes with the host cell, at least two mechanisms could be affected by amino acid substitutions. The first one is the direct or indirect recognition by the resistance protein from the host. Bos et al.  have shown that some mutations leading to specific amino acid substitutions at the position E80K of this protein can be associated with a loss of recognition by the R3a gene product. On the other hand, the replacement of lysine by arginine at the same position of the protein does not affect the cell-death suppression activity of AVR3a . It has been recognized that the variant AVR3aEM produces significantly less hypersensitive response than the other known variants of the gene, producing a virulent phenotype , while the AVR3aKI variant is the most effective in suppressing cell-death in the host . This scenario leads to the possible advantage of maintaining polymorphic residues, which may give the pathogen population different adaptive pathways. The remaining polymorphic positions detected in this study were not under strong positive selection. However, their effect on avirulence on S. tuberosum R3-expressing plants remains to be determined in order to define their potential effect on host resistance at the subspecific level.
Polymorphisms of the Avr3a gene may be related to the plant species from which the isolates were collected. Armstrong et al.  reported that 55 isolates of S. tuberosum from different locations in the world contained only two alleles. In this study we found both previously reported haplotypes AVR3aEM (H1) in S. tuberosum as well as in other Solanum species, and AVR3aKI (H3) restricted to S. muricatum. The other four reported variants were not present in S. tuberosum. Because of this, it is required to establish the correlation of each of these variants and their host range. Additionally, the taxonomic status of the P. infestans- related species P. andina should be clarified in order to understand the real sympatry of the two species and their respective contribution of alleles to the population. Furthermore, a more intensive sampling is needed in order to correlate polymorphisms with the host. Finally, the wider picture of the population structure in this region suggests that the low steps in the networks, together with the low genetic diversity of this pathogen and the evidence of the presence of selection in the effector gene, might be consistent with a scenario of random dispersion of new mutations by drift that in some cases might be fixed by selection. We need further research in other populations and more data from effector genes and neutral loci to really understand the Phytophthora infestans diversification process in South America.