The wild forms of P. vulgaris and P. lunatus are distributed in both Mesoamerica and South America, while those of P. dumosus, P. coccineus, and P. acutifolius have a geographic distribution that is restricted to Mesoamerica (Figure 3). There have been numerous studies that have investigated the origin and evolution of P. vulgaris, which among these five domesticated Phaseolus spp. has the greatest economic importance. In contrast, there have been very few studies into the origins of P. lunatus, P. acutifolius, P. coccineus, and P. dumosus.

Genetic diversity is not uniformly distributed throughout the variety of climatic and environmental conditions under which a species grows. Consequently, it depends not only on demographic processes (e.g., genetic drift), but also on adaptation, in terms of the ability to cope with specific environmental conditions. These conditions can include extremes of cold and heat, lack or excess of water, different light intensities and duration, and diverse soil conditions and pest and disease pressures. As indicated above, the wild forms of Phaseolus crop species grow under a variety of different environmental and climatic conditions, and their geographic distribution mirrors their diverse patterns of adaptation to the different ecological niches, as well as their life histories and reproductive systems (Figure 2). P. vulgaris is adapted to warmer temperatures (mesic and temperate soil temperatures) at lower altitudes, with a rainfall of ∼1,100 mm/year (Supplementary Table S1). P. coccineus is found in more humid environments, at cooler temperatures, and at higher altitudes, while P. dumosus is characterized by intermediate adaptation. P. acutifolius is a drought-tolerant species that originated in warmer and more arid environments (e.g., it is grown in the arid lands of Mexico and in southeast USA). Finally, P. lunatus is particularly suited to low-altitude humid and sub-humid climates, as well as warm temperate zones (Figure 2 and Supplementary Table S1). These observation are confirmed by principal component analysis (Figure 5). This was calculated from the data on 20 ecological variables (Supplementary Table S1) using DIVA-GIS 7.5¹, as inferred from the geographical coordinates of the collection sites of the wild accessions for the passport data present in the database of the International Centre for Tropical Agriculture (CIAT). Moreover, as can be seen in Figure 5, among these five species, P. vulgaris shows the widest ecological adaptation.

There is little in the literature that is aimed at highlighting the genetic basis of adaptation, especially for the wild forms of these crop species, although a few recent studies have been focused on P. vulgaris. Rodriguez et al. (2016) carried out a study on the environmental adaptation of wild P. vulgaris. They investigated the role of demographic processes (e.g., genetic drift) and selection for adaptation in the shaping of the current genetic structure of wild P. vulgaris. This analysis was based on 131 single nucleotide polymorphisms (SNPs) on a sample of 417 wild P. vulgaris accessions that are representative of the geographic distribution of the wild forms of the species. They first investigated the spatial distribution of the genetic diversity of the wild forms of P. vulgaris, using a landscape genetics approach that was based on an individual centered analysis to avoid sampling bias (Figure 6). Briefly, they delineated a circular neighborhood of 100-km radius around each georeferenced accession, and the accessions falling within each neighborhood were used to calculate the relative unbiased gene diversity, He (Nei, 1978). The correlation between He and the neighborhood size was not significant (r = 0.040, n = 299, P = 0.482), and the mean size of each neighborhood was 40.6 individuals and 83.3% of the neighborhoods included >10 individuals (Rodriguez et al., 2016). By using this approach, they observed high genetic diversity across Mexico, from the state of Oaxaca to Durango, with a depression in the genetic diversity in an area that lies approximately across the regions of Guerrero, Morelos, Puebla and Estado de Mexico. Possible explanations for the reduced diversity of this area in Mexico include natural selection due to an environment that is too arid for P. vulgaris, or a genetic bottleneck caused by the volcanic activities that were frequent in this area in ancient times (Márquez et al., 1999; Siebe, 2000; Siebe et al., 2004; Plunket and Uruñuela, 2012). Low diversity was also found in Guatemala, Costa Rica and Colombia, and particularly in the Honduras (Figure 6).

The approach applied by Rodriguez et al. (2016) also highlighted local adaptation at the continental level. The rationale behind the study was that the use of spatial data in combination with genetic diversity data would allow discrimination between the effects of geography and ecology; i.e., demographic processes vs. selection (Bradburd et al., 2013; Wang et al., 2013; Guillot et al., 2014; Kraft et al., 2014). As evidence of the effectiveness of this approach, it was possible to disentangle the effects of geography from ecology in the shaping of the genetic patterns observed, and correlations between markers and ecological variables were detected. By scanning the SNP markers used for the analyses, 26 loci (19.8%) were identified as under signatures of selection, seven of which (5.3%) showed strong probability levels. Although the proportion of loci under selection might be overestimated, as they were chosen among genes that were putatively involved in adaptation, different loci were found to have compatible functions with adaptation features, such as cold acclimation, chilling susceptibility, and mechanisms related to drought stress. Moreover, as well as the well-delineated genetic groups in the Mesoamerican gene pool, they demonstrated global structures for both the neutral loci and the loci under selection. Overall, these data suggest that the origin of the geographic structures might be the outcome of the expansion of the species and gene flow, including crop-to-wild introgression (Papa and Gepts, 2003; Papa et al., 2005), as was also recently confirmed by Rendón-Anaya et al. (2017) using whole genome sequencing analysis. Nonetheless, the adaptation to different environmental conditions might have led to the present population structure, with the subsequent limited long-range gene flow and divergent selection along an ecological cline of variation.

One of the major issues of evolutionary studies that are focused on the domestication of plant species is to identify the geographic areas where they were domesticated (Harlan, 1975).

The population genetics model of domestication predicts a reduction in diversity and increased divergence between wild and domesticated populations, due to demographic factors that affect the whole genome, and because of selection at target loci. Several interesting insights can be revealed by comparisons between different species (Glémin and Bataillon, 2009). There are many examples in the literature that have used different molecular markers and nucleotide data to show the reduction in genetic diversity of crop species compared with their wild progenitors (for review, see Bitocchi et al., 2013). Allogamous species, such as maize (Z. mays), are generally characterized by lower genetic bottleneck effects compared to autogamous species, such as the common bean P. vulgaris, even if other factors can have relevant roles, such as the life history (Bitocchi et al., 2013). Resequencing data have confirmed that in autogamous species, such as soybean (Glycine max) and rice (Oryza sativa, variety japonica) (Lam et al., 2010; Xu et al., 2012), reductions in diversity have arisen due to domestication, as also reported for silkworm and for mammalian species (Xia et al., 2009; vonHoldt et al., 2010; Lippold et al., 2011).

Identification of the genes involved in the domestication process and knowledge of the regions of the genome where those genes are located and of the proportion of the genome affected by domestication are key to better exploitation of the diversity present in the wild relatives, and to enhance the achievements in breeding and crop improvement (Tanksley and McCouch, 1997; McCouch, 2004). As proposed by Cavalli-Sforza (1966) and Lewontin and Krakauer (1973), the identification of loci involved in adaptive processes can be obtained from population genetics expectations that predict that while drift has a homogeneous effect over the genome, selection is acting only for target loci and related linked loci due to the lack of recombination (hitchhiking). Thus, selected (and linked loci) are expected to depart from neutral expectation of diversity and divergence parameters. Moreover, as proposed by Papa et al. (2005), in gene flow between crops and wild forms, aberrant patterns of divergence and diversity can also be determined by the combined actions of asymmetric migration (Papa and Gepts, 2003) and selection at target loci.

Papa et al. (2007) used 2,506 AFLPs for a whole genome scan for the signature of selection due to domestication, and they estimated that about 16% of the genome of the common bean P. vulgaris appeared to be under the effects of selection. Bellucci et al. (2014a) used RNA sequencing technology, and after simulating the demographic dynamics during domestication, they reported that 9% of the genes were actively selected during domestication in Mesoamerica. Furthermore, in these contigs, selection induced a further reduction in the diversity of gene expression (by 26%), and was associated with a fivefold enrichment of the differentially expressed genes.

Bellucci et al. (2014a) also carried out a survey on the function of a subset of contigs that are putatively under selection, to determine whether they are known to be associated with the domestication process in other species, using either direct experimentation or through their function. Interestingly, among the genes putatively under selection that showed greater genetic diversity in the wild compared with the domesticated form, they found sequence homologs to: (i) genes that are involved in ‘light’ response pathway; e.g., GIGANTEA (GI), which has a pivotal role in the photoperiodic response (Mizoguchi et al., 2005; Hecht et al., 2007; Kim et al., 2012); (ii) genes that are pivotal to ensure correct hormonal perception, transport or biosynthesis; (iii) genes that are involved in seed development and traits; and (iv) genes that are involved in responses to environmental stress.

Another interesting example is the homolog of YABBY5 (YAB5), which is a transcription factor that is implicated in the regulation of seed shattering in cereal species, including sorghum (Sorghum bicolor), rice and maize (Lin et al., 2012). In most cases, Bellucci et al. (2014a) found evidence of positive selection associated with domestication, but in a few cases, this selection had increased the nucleotide diversity in the domesticated pool at a target locus associated with abiotic stress responses, flowering time, and morphology. In particular, for 2.8% of the genes putatively identified to be under the effect of selection by Bellucci et al. (2014a), there was no diversity in the wild forms, while there was diversity in the domesticated. They explained this as due to novel mutations (or standing variations) that were selected because of the crop expansion into new environments (diversifying selection) with unexpected biotic and abiotic stress, or because of selection for traits that improved the use of the plant organs by humans (de Alencar Figueiredo et al., 2008). An interesting example was given by the functional analysis of the drought- and growth-related (Osakabe et al., 2013) KUP6 (K⁺ uptake transporter6) gene, where this was significantly overexpressed in the domesticated compared to the wild (Figure 7), as if domestication had also increased the functional diversity of selected genes in addition to the increased nucleotide diversity.

Schmutz et al. (2014) also investigated whether their candidate genes were implicated in important domestication traits, such as flowering time and seed size. A total of 38 flowering genes were identified in the Mesoamerican and Andean candidate lists, while another subset of 15 genes was found to be associated with seed size in genome-wide association studies, and 11 genes contained SNPs that were associated with seed weight.

Recently, Bitocchi et al. (2016a) analyzed nucleotide sequences from a set of 49 gene fragments from a sample of 39 wild and domesticated Mesoamerican accessions of P. vulgaris. In this study, they applied the same approach as Bellucci et al. (2014a), and they identified several loci that showed signatures of selection during common bean domestication in Mesoamerica. In particular, they had the possibility to see if their candidates were detected as outliers also in other studies of varying sizes, data types, and methodologies (Bellucci et al., 2014a; Schmutz et al., 2014; Rodriguez et al., 2016). They thus obtained independent evidence that four genes (i.e., AN-Pv33, AN-Pv69, AN-DNAJ, and Leg223) were targets of directional selection during common bean domestication. The gene function investigation for these genes highlighted that they are involved in plant resistance tolerance to biotic and abiotic stresses, such as heat, drought, and salinity. Moreover, another important outcome of Bitocchi et al. (2016a) was related to the observed excess of non-synonymous mutations in the domesticated germplasm. In particular, they observed a significantly higher frequency of polymorphisms in the coding regions compared to non-coding regions only in the domesticated beans. These mutations were mostly non-synonymous and were recently derived mutations present in genes related to responses to biotic and abiotic stresses. These data cannot be fully explained by the cost of domestication alone, but support a scenario where new functional mutations were selected for adaptation during domestication, showing that domestication also increased the functional diversity at target loci that enable the domesticated forms to successfully compete during the expansion and adaptation to new agro-ecological growing conditions.

The key aspect of domestication is the convergent phenotypic evolution that is associated with the adaptation to a novel agro-ecosystem, and to human needs. For instance, most domesticated animals were selected to maximize the production of useful products (e.g., meat, milk, and wool) and for their docile behavior, while crops were selected for the size of the plant organs used by humans (e.g., seeds and fruit) and for reduced, or lack of, seed dispersal. For these reasons, domestication provides us with a unique tool to understand the process of adaptation, to test evolutionary hypotheses, and to identify the molecular basis of phenotypic diversity. Several interesting insights can be revealed by comparisons among different species (Glémin and Bataillon, 2009), where, for instance, the population genetics model of domestication predicts a reduction in diversity and increased divergence between wild and domesticated populations due to demographic factors that affect the whole genome and because of selection at target loci.

In this regard, an example is given by the study of Nanni et al. (2011), in which wild and domesticated P. vulgaris accessions were analyzed by sequencing a genomic region of ∼1,200 bp (PvSHP1) that is homologous to SHATTERPROOF-1 (SHP1), a gene involved in the control of fruit shattering in Arabidopsis thaliana (Liljegren et al., 2000). The loss of fruit shattering has been under selection in most seed crops, to facilitate seed harvesting (Purugganan and Fuller, 2009), while in wild plants, this feature is a fundamental trait enabling seed dispersal. Expressed sequences that correspond to SHP1 have also been identified in other species, such as, tomato, where it was indicated as having an important role in the regulation of both fleshy fruit expansion and the ripening process (Vrebalov et al., 2009), which are together necessary to promote seed dispersal of fleshy fruit. In legumes, sequences orthologous to Arabidopsis SHP1 have been identified in Medicago, pea and soybean (Hecht et al., 2005). Nanni et al. (2011) mapped PvSHP1 on linkage group Pv06 of the common bean genome, and showed that it did not co-segregate with the St locus, which is responsible for the presence or absence of pod string, and was mapped on linkage group Pv02 using a domesticated (Midas) × wild (G12873) RIL population by Koinange et al. (1996). These results suggested that PvSHP1 is not responsible for the observed phenotypic variation in P. vulgaris for fruit shattering. Similar results were found by Gioia et al. (2013b), who sequenced and mapped PvIND on the common bean genome, a sequence that is homologous to the INDEHISCENT gene (IND), which is the primary factory required for silique shattering in A. thaliana. PvIND mapped near the St locus; however, the lack of complete co-segregation between PvIND and St and the lack of polymorphisms at the PvIND locus correlated with the dehiscent/ indehiscent phenotype suggested that PvIND is not directly involved in pod shattering and is not the gene underlying the St locus (Gioia et al., 2013b).

Recently, Murgia et al. (2017) developed a phenotyping approach in P. vulgaris to evaluate the shattering syndrome in a segregating population. This is a promising approach for the identification of genetic factors that control the shattering trait in the common bean, and it will greatly facilitate comparative studies among legume crops, and also gene tagging.

Within the same species, the study of Schmutz et al. (2014) represents the first example of the possibility to investigate convergent evolution between the two gene pools of the common bean P. vulgaris. Indeed, a comparison of the results of selection in the two gene pools, in which independent domestications occurred, allowed them to determine whether to obtain the same convergent phenotypes, evolution took part in the selection of the same genomic regions or of completely different set of genes that code for the same phenotypes. Interestingly, only a small portion of the genome and of genes identified as putatively under selection during domestication were shared between the two gene pools, which suggested different genetic routes to domestication (Schmutz et al., 2014). This outcome appears to suggest that the sexually compatible Mesoamerican and Andean lineages with similar morphologies and life cycles underwent independent selection upon distinct sets of genes. However, taking into account that explicit demographic modeling was not used to generate an expectation of the number of potential false positive regions by Schmutz et al. (2014), another possible explanation for this result is that the lack of correlation between the two gene pools is due to a high level of false positives; i.e., regions of the genome with reduced diversity due to the stochastic effects of domestication bottlenecks. Regarding this consideration, Bitocchi et al. (2016a) compared their candidates for selection during domestication of common bean in Mesoamerica with those of other studies (Bellucci et al., 2014a; Schmutz et al., 2014; Rodriguez et al., 2016), and found that two (AN-Pv69, AN-DNAJ) out of the four strong candidates identified were detected as outliers by Schmutz et al. (2014) only during Andean domestication. This implies that more studies are needed either to support or refute the lack of correlations between the two gene pools found by Schmutz et al. (2014).

High levels of genetic diversity have been reported for common bean populations cultivated worldwide, and several continents and countries have been proposed as secondary centers of diversification for this species. These have included: the Iberian Peninsula (Santalla et al., 2002), the whole of Europe (Angioi et al., 2010, 2011; Gioia et al., 2013a), Brazil (Burle et al., 2010), central-eastern and southern Africa (Martin and Adams, 1987a,b; Asfaw et al., 2009; Blair et al., 2010), and China (Zhang et al., 2008).

In South America, a particular situation has emerged in Brazil. Although Brazil is closer to the Andes than to Mesoamerica, unexpectedly, it is the Mesoamerican P. vulgaris that is more prevalent (Burle et al., 2010). Multiple introductions of Mesoamerican germplasm in periods antecedent or successive to the discovery of the Americas might explain this pattern (Gepts et al., 1988).

In Africa, overall, the two gene pools are approximately equal in frequency, albeit there are strong differences between countries (Angioi et al., 2011; Okii et al., 2014). Such differences have been explained by the existence of at least partially independent seed networks in different countries (Asfaw et al., 2009), and because of selection due to dissimilar ecological and economic conditions among countries (Wortmann et al., 1998; Asfaw et al., 2009). Differential resistance to soil-borne diseases like Fusarium root rot, and different yield performances arising from the ‘interference’ of improved genotypes released by national breeding programs have also been considered to explain the uneven distribution of the two gene pools across these regions (Blair et al., 2010).

In China, a prevalence of the Mesoamerican P. vulgaris has been observed (Zhang et al., 2008), although this was attributed mainly to founder effects (Zhang et al., 2008). The Himalayan region, as also the entire Indian subcontinent, shows high genetic diversity (Sofi et al., 2014; Rana et al., 2015).

Little is known about the dissemination of the common bean in India, albeit trading in the 16th century via the Red Sea and the Arabian Sea, and through the Hindustan Silk route, probably had a determinant role in the dissemination of this crop. It is also possible that the sea route discovered by the Portuguese explorer, Vasco da Gama, had a role in this dissemination. The genetic diversity in India also includes the combination of both the Mesoamerican and the Andean gene pools (Rana et al., 2015). Moreover, adaptation to micro-geographic conditions has been suggested for these landraces. Indeed, the analysis of >4000 landraces allowed the identification of several diverse clusters, irrespective of the place of collection, which also indicates a strong role for gene flow.

The Mesoamerican and Andean gene pools were both introduced into Europe. Studies carried out using several different marker types have shown that the Andean P. vulgaris predominates over Mesoamerican P. vulgaris (Gepts and Bliss, 1988; Lioi, 1989; Santalla et al., 2002; Logozzo et al., 2007; Angioi et al., 2010). The Andean type is largely predominant for the Iberian peninsula, Italy and central-northern Europe, where it is also prevalent on a local scale (Sicard et al., 2005; Angioi et al., 2009b). In the eastern part of Europe, the frequency of the Mesoamerican type tends to increase but it is always lower than the Andean (Papa et al., 2006). In their study of the expansion of the common bean P. vulgaris in Europe, Angioi et al. (2010) concluded that the intensity of the cytoplasmic bottleneck that resulted from this introduction into Europe was very low or absent (i.e., a loss of cpSSR diversity of ∼2%).

Regarding the other Phaseolus crop species, there is little in the literature that has focused on the investigation of genetic diversity of P. coccineus, the allogamous sister species of P. vulgaris, in Europe. Some studies were conducted on small (Nowosielski et al., 2002; Sicard et al., 2005; Acampora et al., 2007; Boczkowska et al., 2012) and ample (Spataro et al., 2011; Rodriguez et al., 2013) spatial scales. The introduction of P. coccineus into Europe was probably contemporary with that of the common bean P. vulgaris (Westphal, 1974). Among the Mediterranean countries, P. coccineus is more widespread in Spain and Italy, while in northern Europe, it occurs more often in the UK and The Netherlands, where P. coccineus has often substituted for P. vulgaris (Santalla et al., 2004) as it is more adapted to cold temperatures and cool summers than P. vulgaris (Delgado Salinas, 1988; Rodino et al., 2007).

However, overall, P. coccineus is characterized by a narrower adaptability to environmental conditions than P. vulgaris. As previously mentioned, P. coccineus and P. vulgaris are cross-fertile when P. vulgaris is the maternal parent, and this might have allowed hybridisation between these two species in Europe, where they often coexist in close sympatry in the same field. However, no evidence of introgression between common bean and runner bean has been found (Sicard et al., 2005). Data on nuclear and chloroplast variability of P. coccineus indicate that in Europe it has at least two main genetic groups (Spataro et al., 2011; Rodriguez et al., 2013). Of particular interest, there is a highly significant association between latitude and phenology for P. coccineus (Rodriguez et al., 2013). This relationship still holds when the effects of population structure for cpSSRs and nuSSRs is factored out. Therefore, this correlation is not just a consequence of the uneven geographic distribution of the two P. coccineus gene pools across Europe. It was then suggested that selection (probably for photoperiod sensitivity, and/or for low temperature), rather than migration and gene flow, has also had a role in shaping the population structure of P. coccineus in Europe (Rodriguez et al., 2013).

A comparison between Spanish and Mexican accessions of P. vulgaris and P. coccineus suggested that P. coccineus has maintained a high level of diversity since its introduction into Europe (Alvarez et al., 1998). A more recent study analyzed a worldwide collection of P. coccineus, and this analysis indicated that limited diversity of the runner bean P. coccineus appears to have been introduced into Europe, and that for nuclear markers, the European landraces show a reduction in diversity of 33% compared to that of the Mesoamerican landraces (Spataro et al., 2011). More recently, the use of chloroplast markers indicated a moderate-to-strong cytoplasmic bottleneck that followed the expansion of P. coccineus into Europe, with a reduction of 13% in chloroplast diversity (Rodriguez et al., 2013). As these markers are the same in number and type as those used by Angioi et al. (2010) to estimate the bottleneck of P. vulgaris following its expansion into Europe (2%), it can be concluded that the loss of diversity in P. coccineus appears to be stronger than in P. vulgaris.

Both nuclear and chloroplast analyses have shown that Mesoamerican and European P. coccineus accessions belong to distinct gene pools (Spataro et al., 2011; Rodriguez et al., 2013). It can be hypothesized that the differentiation of the European gene pool was due to adaptation to the new environment, and to genetic drift and a lack of introgression from wild forms.

The introduction of P. lunatus in Europe was very limited, with very few examples (Doria et al., 2012).

Using isozyme markers, Santalla et al. (2002) estimated a high percentage of hybrids in their common bean landrace collection from the Iberian peninsula (25%). Then later, through the integration of cytoplasmic and nuclear analyses, Angioi et al. (2010) reported that about 44% of their wide collection of landraces that spanned almost all of the European countries appeared to be derived from at least one hybridisation event. In addition to the molecular results, Angioi et al. (2010) showed that seed size and coat traits vary with the level of introgression between the two gene pools. More recently, Gioia et al. (2013a) used population assignment techniques to also reveal extensive hybridisation between the two gene pools in Europe, with a frequency of hybridisation that was almost fourfold greater in Europe (40.2%) than in the Americas (12.3%), which confirmed the findings of Angioi et al. (2010). This can be explained by the geographic isolation between the gene pools in the American centers of origin, and that following the introduction into Europe, genotypes from different genes pools often coexisted on very small spatial scales (i.e., in small cultivated areas), and thus had the chance to hybridize. Estimations of hybridisation in other parts of the world have always been <10%, and so much less than in Europe (Zhang et al., 2008; Asfaw et al., 2009; Blair et al., 2010; Burle et al., 2010). Taken all together, this evidence supports the hypothesis that the whole of Europe can be regarded as a secondary center of diversity for the common bean (Angioi et al., 2010), as also suggested by the work of Santalla et al. (2002) that was limited to the Iberian peninsula.

After its introduction into Europe, P. vulgaris was exposed to new and ample agro-ecological conditions. Thus, it is likely that any releases that were not adapted to the new conditions were initially purged by selection. Among the natural factors, biotic and abiotic stress were probably determinants in shaping the genetic structure of the European bean landraces. It is possible that their adaptation to long days, their cold tolerance, and their resistance to pests and diseases were crucial, and this would probably have led to a reduction in the diversity that was initially present in the founding populations. Additionally, the selection operated by farmers for seed color and size, and culinary, organoleptic and nutritional quality might also have had strong impact on the evolution of the European bean, as witnessed by the myriad of local bean populations with particular characteristics and specific names (Angioi et al., 2009b; Lioi and Piergiovanni, 2013; De Ron et al., 2016). Moreover, the documented scenario of extensive Mesoamerican × Andean gene pools through their hybridisation in Europe (Logozzo et al., 2007; Angioi et al., 2010; Gioia et al., 2013a) suggests that introgressive hybridisation might have been the fundamental ‘evolutionary stimulus’ (Anderson and Stebbins, 1954; Lewontin and Birch, 1966) that propelled and boosted bean evolution in Europe. Indeed, hybridisation can produce new genotypic and phenotypic combinations that do not occur in either of the parental taxa, and upon which selection might act.

Recently, Bitocchi et al. (2015) reported that European flint landraces grown in situ show adaptive introgression from modern maize. A key result of their study was that adaptation followed by hybridisation has been very rapid, with landraces capturing and increasing the frequency of favorable alleles over very short times (e.g., 50 years) (Bitocchi et al., 2015). This allows the hypothesis that this evolutionary mechanism might also have operated for the European bean landraces, which have a history that is some 10-fold longer. This is of interest not only for studies in evolutionary genetics, but also for plant breeders. Indeed, studies focused on hybridisation have shown the potential for the identification of functionally important regions of the genome (Arnold and Martin, 2010).

The potentiality of the studies at continental scale on the architecture of adaptation in common bean is confirmed by the possibility to apply a signature of selection approach using European landraces. Indeed, using methods for the identification of outlier loci for selection, Santalla et al. (2010) provided evidence that selective forces have had significant roles, particularly for seed size, growth habits, pest resistance and flowering time. This is intriguing, as selection for flowering time was probably a key element in the history of the bean also in its center of origin.

The BEAN_ADAPT project² is funded through the 2nd ERA-CAPS call, ERA-NET for Coordinating Action in Plant Sciences. The main aim of this project is to dissect out the genetic basis and phenotypic consequences of the adaptation to new environments of P. vulgaris and its sister species P. coccineus, through the study of their introduction from their respective centers of domestication in the Americas and their expansion through Europe as a recent and historically well-defined event of rapid adaptation. BEAN_ADAPT thus plans to characterize a large collection (11,500 accessions of each species) from three major genebanks by genotyping-by-sequencing. This will define the population structure and obtain subsets of genotypes for phenotyping (e.g., field, growth chamber) and for deeper genomic–transcriptomic–metabolomic characterisation. A multidisciplinary approach is planned (i.e., genomics, population/quantitative genetics, biochemistry, plant physiology) to unraveling the genetic bases of adaptation of these crops in new agro-ecological environments. The methods here rely on previous studies that have demonstrated the effectiveness of various analysis, such as the application of: (i) population genomics, to test for signatures of selection (e.g., Bellucci et al., 2014a); (ii) evolutionary metabolomics, which appear to be a very powerful approach to characterize molecular phenotypic changes due to domestication, and to identify traits under selection in wheat (Beleggia et al., 2016); (iii) association mapping studies of complex traits, such as flowering time (Raggi et al., 2014); (iv) analysis of signatures of selection by searching for ‘unusually high’ correlations between SNPs and environmental variables at the continental scale, as successfully applied by Rodriguez et al. (2016) for wild common bean from Mexico.