Introduction
Evidence obtained during recent years suggests that the common bean (Phaseolus vulgaris L.) was domesticated in Mesoamerica and the Andes, but its secondary centre of genetic diversity probably extended to Brazil, China and Europe (Chacón et al., Reference Chacón, Pickersgill and Debouk2005; Zhang et al., Reference Zhang, Blair and Wang2008; Burle et al., Reference Burle, Fonseca, Kami and Gepts2010; Santalla et al., Reference Santalla, De Ron and De La Fuente2010). After domestication, this species has become one of the most important crop plants in developing countries, as it is an economical source of important nutritional components (Santalla et al., Reference Santalla, Fueyo, Rodino, Montero and De Ron1999).
Nicaragua is located in the Mesoamerican centre of genetic diversity for Phaseolus species, where a high genetic diversity is expected to prosper in diverse environmental conditions. Some studies have aimed to describe the genetic diversity content in a small number of Nicaraguan landraces (Gómez et al., Reference Gómez, Blair, Frankow-Lindberg and Gullberg2004, Reference Gómez, Blair, Frankow-Lindberg and Gullberg2005). However, most of the genetic diversity remains without any estimation and is undervalued as a potential source of genetic variation. On the other hand, many landraces and old cultivars that were quite popular some years ago are in danger to become extinct and some of them, according to recent expeditions, have already been lost (Oswalt R. Jiménez, personal observation). Thus, a proper estimation of the genetic diversity of on-farm conserved landraces is urgently needed. They have evolved along with farmers' preferences, as their subsistence relies on the beans' good adaptation capacity and high culinary quality, traits that are appreciated in national and international markets. Nicaraguan landraces produce very low yields (Gómez and Frankow-Lindberg, Reference Gómez and Frankow-Lindberg2005) when compared with improved lines and cultivars, and this characteristic is the main disadvantage for producing these beans. However, despite yield being influenced by the environment, this trait can be improved if proper levels of genetic variation are present in breeding programmes (Falconer and Mackay, Reference Falconer and Mackay1996; Acquaah, Reference Acquaah2007).
The main bean cultivars that are produced in Nicaragua nowadays were improved by regional bean breeding programmes using germplasm from different sources (Jiménez, Reference Jiménez2009). These programmes produce new cultivars for the Central American region where the preferences and environmental conditions vary among different countries. After validating the promising genetic material in Nicaraguan conditions, most advanced lines do not meet the requirements (distinctness, uniformity and stability) and market acceptance, and only a very few lines are finally released as new cultivars.
The identification of novel sources of genetic variation and their use in local breeding programmes can justify and further enhance the conservation of locally adapted bean genetic resources in countries where a robust conservation strategy is still missing. The objectives of the present study were to investigate the genetic diversity of Nicaraguan common bean landraces using microsatellite markers and to identify promising sources of genetic variation for breeding purposes.
Materials and methods
Germplasm collection and population selection
Between March and May 2010, four expeditions were carried out to different regions of Nicaragua with the aim to establish a seed bank. When visiting the farms, small amounts of seeds were requested from the farmers and passport data were recorded. From the information obtained, the reasons why farmers prefer and produce those populations were especially valued in the registers. We sampled each bean population by taking at least 300 seeds from different bean bags, respectively, in order to have a representative sample. The amount of seeds collected from each population varied from 300 to 1000 g (depending on seed availability). The seeds were cleaned and their physiological qualities were tested at the National Center of Agricultural Research and Biotechnology (CNIAB). The 100-seed weights, among other seed traits, were determined following the ISTA rules (ISTA, 2004). Seed length and width were determined as in Blair et al. (Reference Blair, Diaz, Buendia and Duque2009). The accessions were then conserved in a cold room (12°C) at the CNIAB. Geographic origin, diverse seed coloration and high level of acceptance by farmers were the criteria for selection. Forty out of 200 bean accessions collected during the expeditions, including three Tepary bean (Phaseolus acutifolius A. Gray) populations, were chosen for this study (Table 1). According to farmers' statements and an information review, all chosen populations have never been studied before. Thus, they represent a novel source of information regarding Nicaraguan bean genetic resources. Additionally, breeder's seeds from two cultivars, ‘INTA ROJO’ and ‘INTA FUERTE SEQUIA’, were included as reference populations.
Forty bean landraces selected from 200 accessions in an ex situ collection, their seed features and origin in Nicaragua
a The highest 100-seed weights were found in cultivars ‘INTA ROJO’ (24.5 g) and ‘INTA FUERTE SEQUIA’ (27.3 g). Seed lengths and widths averaged 1.04 and 0.61 cm, respectively.
b Apparently an old cultivar.
c Possibly a species different from Phaseolus vulgaris L.
d Tepary bean (Phaseolus acutifolius A. Gray).
DNA extraction
DNA extraction from individual seedlings (germinated in sterile sand), ten randomly sampled individuals per accession, was carried out at the Biotechnology Laboratory at the CNIAB in Nicaragua. The mini-preparation protocol (Dellaporta et al., Reference Dellaporta, Wood and Hicks1983) was modified to be used in common beans. About 15 mg of leaf tissue was taken from each individual and placed into a sterile Eppendorf tube. Thereafter, 200 μl of cold miniprep II extraction buffer (containing 100 mM Tris–HCl, 50 mM EDTA, 500 mM NaCl and 20 mM 2-mercaptoethanol) was added, and the tissue was macerated and homogenized using a plastic pestle. Then, 26 μl of sodium dodecyl sulphate (10%) was added and mixed. Subsequently, the samples were placed into a water bath (65°C) for 10 min. Then, 120 μl of 3 M potassium acetate was added, and the samples were incubated on ice for 25 min and centrifuged at full speed (14,800 g ) for 15 min. Then, 200 μl of the supernatant was removed carefully and placed into a clean Eppendorf tube avoiding the debris. Afterwards, 120 μl of extra-pure isopropanol was added, and the samples were kept at − 24°C for 30 min and centrifuged at full speed (14,800 g ) for 15 min. Then, isopropanol was removed, and the pellet was washed twice with ethanol (70%) and left to dry. The pellet was dissolved in 100 μl of TE (containing 10 mM Tris–HCl, pH 8.0 and 1 mM EDTA). Finally, the samples (ten individual DNA samples per accession, 420 in total) were placed into a freezer at − 24°C.
Microsatellite genotyping
Microsatellite genotyping was carried out in the laboratory of the Department of Agricultural Sciences, University of Helsinki, Finland. The following 14 microsatellite markers were selected for this study: BM205, AG1, BM154, BM156, BM184, BM189, BM210, BM212, BM114 (Gaitán et al., Reference Gaitán, Duque, Edwards and Tohme2002), PVag001, PVag004 (Yu et al., Reference Yu, Park, Poysa and Geps2000), BMd8, BMd53 (Blair et al., Reference Blair, Pedraza, Buendia, Gaitán-Solís, Beebe, Gepts and Tohme2003) and ATA10 (Blair et al., Reference Blair, Buendía, Giraldo, Métais and Peltier2008). The marker selection was mainly based on the high level of polymorphisms reported in previous studies and the markers' wide distribution in the common bean genome. Also, these markers have been associated with important QTLs (Quantitative trait loci) for yield components when assayed in materials from different genetic backgrounds (Blair et al., Reference Blair, Iriarte and Beebe2006b, Reference Blair, Diaz, Buendia and Duque2009, 2010; Rodrigues et al., Reference Rodrigues, Dos Santos, Ramalho, Amorim and Silva2007; Pereira et al., Reference Pereira, Dos Santos, De Souza and Lima2008; Torga et al., Reference Torga, Dos Santos, Pereira, Furtado and Leite2010).
Polymerase chain reactions (PCRs) were carried out in 10 μl volumes by mixing the following components: 1 μl of 10 × buffer, 0.2 μl of dNTPs (10 mM each), 6 μl of MQ water (Millipore Quality water), 1 μl of each primer (5 pmol, forward primers fluorescently labelled with FAM (6-carboxyfluorescein), TET (Tetrachlorofluorescein) or HEX (Hexachlorofluorescein) labels), 0.3 μl of DNA polymerase (Dynazyme, 2U/μl) and 0.5 μl DNA template (about 30 ng). The PCRs were carried out as follows: DNA denaturation at 94°C for 4 min followed by 30 cycles of denaturation at 94°C for 45 s, annealing at 46–58°C (depending on the primer pair) for 45 s, and elongation at 72°C for 1 min, with a final elongation at 72°C for 10 min. After amplification, the PCR products were diluted with MQ water at the 1:200 ratio. Of this solution, 0.5 μl of each PCR product was mixed with 20 μl of HiDi-formamide and 0.15 μl of size standard (GeneScan 500 ROX). After mixing, the samples were denatured for 5 min at 95°C. Finally, DNA fragments were analysed using a capillary electrophoresis system (3730 DNA Analyzer; Applied Biosystems) in the Sequencing Laboratory of the Institute of Biotechnology, University of Helsinki, Finland. The allele sizes were determined using the software Peak Scanner version 1.0 (Applied Biosystems).
Statistical analyses
For all populations, the distribution of genetic variation was revealed by an analysis of molecular variance (AMOVA), and the observed and expected heterozygosities (H obs and H exp) were determined using the software Arlequin version 3.5.1.2 (Excoffier and Lischer, Reference Excoffier and Lischer2010). Genetic diversity parameters were estimated for a set of populations (total number of alleles, alleles per locus, number of genotypes and F IS index), and allele frequencies were calculated and compared using Fisher's exact test with software Genepop version 4.1 (Rousset, Reference Rousset2008). To estimate pairwise differences among populations, F ST values were calculated using the software Arlequin version 3.5.1.2 (Excoffier and Lischer, Reference Excoffier and Lischer2010). Then, the pairwise matrix was clustered by the unweighted pair group method of arithmetic averages (UPGMA), and the phylogenetic tree was plotted using the software MEGA version 5.1 (Tamura et al., Reference Tamura, Peterson, Peterson, Stecher, Nei and Kumar2011).
Subsequently, we conducted a Bayesian analysis of population structure with the software Structure version 2.3.3 (Pritchard et al., Reference Pritchard, Stephens and Donnelly2000). We applied an admixture and correlated allele frequencies model where individuals may have mixed ancestry and allele frequency at each locus is correlated along the populations. To determine the number of clusters (K value), we first tested a continuous series of Ks (1–12) in five independent runs for each K value, with a length of burn-in period of 10,000 iterations followed by 100,000 MCMC (Markov chain Monte Carlo) iterations. The five log-likelihood values for each K were then charted to infer the K values around a plateau of the likelihood values. The identified candidate K values were further tested in ten independent runs for each K with a burn-in period of 100,000 iterations followed by 1,000,000 iterations. The most likely K value was determined by analysing ΔK values using the method proposed by Evanno et al. (Reference Evanno, Regnaut and Goudet2005). Populations with a proportion of membership less than 0.8 were considered putative hybrids (Santalla et al., Reference Santalla, De Ron and De La Fuente2010).
Additionally, considering that seed features (length, width and weight) are the only quantitative traits available from the collection, we implemented a germplasm-regression-combined marker–trait association to test specific genotype associations with these traits as proposed in Ruan et al. (Reference Ruan, Li and Mopper2009) and Ruan (Reference Ruan2010). The Tepary bean accessions and the population PV0037 were excluded from these analyses, because the former represents another species and the latter possibly belongs to a different genetic background (based on seed features). Basically, quantitative traits, seed length, width and weight, were treated as dependent variables as implemented by Virk et al. (Reference Virk, Ford-Lloyd, Jackson, Pooni, Clemeno and Newbury1996). All microsatellite genotypes were scored into a binary dataset (1 for presence or 0 for absence) and then considered as independent variables. Stepwise multiple regression analysis (MRA) was based on the following model:
where Y represents quantitative traits; m j represents marker genotypes; b j represents partial regression coefficients that specify empirical relationships between Y and m j ; d represents population residuals; and e is the random error of Y that includes environmental variation (Virk et al., Reference Virk, Ford-Lloyd, Jackson, Pooni, Clemeno and Newbury1996). F values with P values between 0.045 and 0.099 were employed to enter and remove independent variables from the regression equation, respectively (Ruan et al., Reference Ruan, Li and Mopper2009; Ruan, Reference Ruan2010). R 2 denotes the coefficient of determination. Multiple regression analyses were carried out using the software SPSS version 16.0 (SPSS, Chicago, IL, USA; http://www.spss.com). After the regression analyses, the selected marker genotypes were tested with linear curve fitting using linear models to confirm the significance of Beta statistics (β) for each genotype identified by the MRA.
Results
Allelic and genotypic diversity
The microsatellite markers produced good PCR products, except for the loci BM205 and ATA10 that did not produce any PCR product in Tepary bean accessions. The loci BM205 and PVag004 produced a multi-banding pattern (two and between four and six DNA fragments, respectively). As the locus PVag004 produced a complex banding pattern, it was excluded from the genetic analyses. One of the amplified fragments produced by the locus BM205 was monomorphic in the whole array of populations and, consequently, also excluded from the data analyses. AMOVA showed that most genetic variation (64.3%) was present among the populations, while 35.7% of the variation occurred within them. The thirteen microsatellite markers revealed 115 different alleles in total. The average of alleles per locus was 8.9, ranging from 2 (BMd53) to 27 (BM154). A total of 134 genotypes in all populations were identified (Table 2). The highest number of genotypes was displayed at the locus BM154 (29 genotypes). The locus PVag004 producing multi-banding patterns amplified ten different DNA fragments (96, 98, 104, 198, 200, 202, 204, 238, 240 and 242 bp). When the allele frequencies of all populations were compared, Fisher's exact test revealed highly significant differences in frequencies at all loci (P = 0.000). Once the average H exp and H obs values were contrasted, it was evident that the observed heterozygosities were very low (0.034) compared with the expected heterozygosities (0.423). The average F ST value was 0.625 and the F IS value was estimated to be 0.914. Overall, when the number of alleles detected in this study was compared with the results obtained when the markers were reported for the first time, five out of 13 markers displayed a higher number of alleles (Table 2).
Summary of allele and genotype data obtained from the 14 microsatellite loci in 42 bean populations
a Alleles reported in Yu et al. (Reference Yu, Park, Poysa and Geps2000), Gaitán et al. (Reference Gaitán, Duque, Edwards and Tohme2002) and Blair et al. (Reference Blair, Pedraza, Buendia, Gaitán-Solís, Beebe, Gepts and Tohme2003).
b Complex banding pattern, thus excluded from the analyses.
When populations were assessed individually, landraces PV0006, PV0013, PV0028, PV0023 and PV0024 were found to possess the highest amount of allelic variation, excluding Tepary bean populations and PV0037, ranging from 27 to 31 alleles (mean 2.4 alleles per polymorphic locus). When the four most polymorphic markers (BM156, BM154, BM184 and BM114) were scored, populations PV0006, PV0013, PV0023, PV0024 and PV0027 averaged from three to four alleles per locus. On the other hand, landraces PV0005, PV0031 and PV0035 proved unrelated to other populations, i.e. they possessed allele combinations not shared with other populations.
Genetic structure
The UPGMA tree derived from the pairwise F ST values showed that there is moderate genetic differentiation among the populations. The Tepary bean populations, PA0001, PA0003 and PA0002, were clearly plotted as an outgroup in the tree. In the same way, the population PV0037 displayed a different microsatellite profile when compared with the other populations (Fig. 1). The genetic distances suggest that the landrace PV0006 is closely related to the cultivar ‘INTA FUERTE SEQUIA’. In general, the phylogenetic clustering of the whole set of populations does not reflect the origin of the populations, as populations from different agroecological regions could be located in the same phylogenetic branch of the tree.
UPGMA tree describing the genetic relatedness among the 40 common bean landraces and two cultivars from Nicaragua (left), and estimated population structure (K = 3) sorted by the membership coefficient (Q value) in correspondence to phylogenetic inference (right). Cluster 1 contains Phaseolus acutifolius populations and landrace PV0037, cluster 2 contains populations with relatively large seeds and cluster 3 contains populations with relatively small seeds on average. IR = ‘INTA ROJO’; IFS = ‘INTA FUERTE SEQUIA’. A colour version of this figure can be found online at http://www.journals.cambridge.org/pgr.
The Bayesian analyses confirmed the phylogenetic tree clustering, capturing most of the genetic diversity into three groups (K = 3). The cluster membership was assigned as follows: cluster 1 involved Tepary bean populations (PA0001, PA0002 and PA0003) and the population PV0037. Clusters 2 and 3 contained common bean populations, 15 out of 38 populations belong to cluster 1 (39.5%) and 23 out of 38 to cluster 3 (60.5%). Cluster 2 comprised a slightly higher allelic diversity (on average 22.5 alleles per population across the 13 loci) compared with cluster 3 (21.4 alleles). Most genetic variation in these clusters is ascribed to within-population variation (69.7%) with a F ST value of 0.321, as revealed by AMOVA. Cluster 2 included two reference cultivars, ‘INTA ROJO’, ‘INTA FUERTE SEQUIA’ and the landraces PV0004, PV0005, PV0006, PV0007, PV0016, PV0017, PV0018, PV0021, PV0024, PV0026, PV0027, PV0029 and PV0031. Finally, cluster 3 encompassed landraces PV0001, PV0002, PV0003, PV0008, PV0009, PV0010, PV0011, PV0012, PV0013, PV0014, PV0015, PV0019, PV0020, PV0022, PV0023, PV0025, PV0028, PV0030, PV0032, PV0033, PV0034, PV0035 and PV0036 (Fig. 1). The populations PV0015, PV0020, PV0024, PV0028, PV0030, PV0034 and PV0036 possessed a coefficient of membership less than 0.8 and were considered as hybrid populations under admixture structure. The proportion of non-hybrid populations was 83.3%. As common bean populations were inferred into two clusters, we conducted an AMOVA omitting populations from cluster 1. These results showed that 47.7% of the variation represents within-population variation and 53.3% among-population variation.
Association of microsatellite markers with seed features
Stepwise MRA was conducted to outline the correlation of 107 microsatellite genotypes with the seed length, width and weight of 38 common bean populations. Three stepwise runs were programmed in order to get the best independent variables that explain the variation. After discarding unsuitable and hybrid genotypes, six alleles from four microsatellite markers (BM205, AG1, BM156 and PVag001) explained most of the phenotypic variation (Table 3). The associations of alleles with seed weight were also tested with a curve-fitting programme, which confirmed a linear relationship.
Results of stepwise multiple regression analyses conducted for microsatellite genotypes associated with seed features
MRA, multiple regression analysis.
a Within each trait, different parameters were estimated when genotypes were included in succeeding MRA steps.
b Subscripts correspond to fragment sizes (bp) for each locus.
c R 2 is the coefficient of determination, which expresses the amount of variation explained by the independent variable.
d R 2 change is the change in R 2 statistics.
e F change is the change in F statistic that are produced when an independent variable is added or deleted.
For seed length, allele 138 from the marker AG1 (called as AG1 138 ), allele 224 from the marker BM156 (called as BM156 224 ) and allele 278 from the marker PVag001 (called as PVag001 278 ) showed a correlation with seed length (Table 3). Genotypes AG1 138 and PVag001 278 showed negative correlations and genotype BM156 224 showed a positive correlation. Genotype PVag001 278 showed the highest (R 2 = 0.214) significant (P = 0.003, t = − 3.133) correlation with a high standardized β value of − 0.463. When the genotypes BM156 224 and AG1 138 were added to the model, the correlation increased (R 2 = 0.339 and 0.423, respectively). For seed width, only allele 222 from the marker BM156 (called as BM156 222 ) showed a positive (R 2 = 0.175) significant (P = 0.009, t = 2.767) correlation with this trait, with a standardized β value of 0.419.
For seed weight, two alleles from the marker BM205, alleles 131 and 133 (called BM205 131 and BM205 133 ) showed significant correlations with seed weight (Table 3; Fig. 2). The genotype BM205 131 showed the greatest (R 2 = 0.340) highly significant (P = 0.000, t = 4.304) positive correlation with seed weight. The standardized β coefficient was also high (0.583). When both genotypes were included in the model, the correlation increased (R 2 = 0.429). Information concerning correlations, regressions and ANOVAs for these four markers is presented in Table 3.
Regression (P = 0.000) plotted for seed weight and the frequency of the microsatellite genotype BM205 131 in 38 common bean populations (see Table 1 for population information). Black squares represent the five populations possessing the smallest seeds (36, 35, 1, 3 and 19), and grey squares represent the five populations with the largest seeds (38, 37, 27, 7 and 5). IR = ‘INTA ROJO’; IFS = ‘INTA FUERTE SEQUIA’.
Discussion
In the present study, we investigated the genetic diversity of 40 common bean landraces from Nicaragua and two cultivars using microsatellite markers, and identified promising populations for breeding purposes. The detected level of genetic variation (mean 8.9 alleles per locus) was higher than that previously reported for small red-seeded landraces (the same market class as studied here) in Nicaragua, 5.7 and 4.3 alleles per locus (Gómez et al., Reference Gómez, Blair, Frankow-Lindberg and Gullberg2004, Reference Gómez, Blair, Frankow-Lindberg and Gullberg2005), and higher than that reported in most other studies on common beans where microsatellite markers have been used (Blair et al., Reference Blair, Giraldo, Buendía, Tovar, Duque and Beebe2006a; Díaz and Blair, Reference Díaz and Blair2006; Benchimol et al., Reference Benchimol, De Campos, Morais, Colombo, Chioratto, Fernandes, Lima and Pereira2007; Zhang et al., Reference Zhang, Blair and Wang2008; Díaz et al., Reference Díaz, Buendía, Duque and Blair2010; Santalla et al., Reference Santalla, De Ron and De La Fuente2010; Cabral et al., Reference Cabral, Soares, Lima, De Miranda, Souza and Gonçalves2011; García et al., Reference García, Rangel, Brondani, Martins, Melo, Carneiro, Borba and Brondani2011; Avila et al., Reference Avila, Blair, Reyes and Bertin2012). Typically, lower levels of genetic variation have been detected in common bean populations when compared with other self-pollinated species (Santalla et al., Reference Santalla, De Ron and De La Fuente2010). A very high degree of genetic diversity has been discovered by Blair et al. (Reference Blair, Diaz, Buendia and Duque2009), on average 18.4 alleles per locus. However, this study was conducted on a core collection holding accessions from different species, centres of origin and races. In more standard microsatellite comparisons, averages ranging from 2.8 to 7.8 have been found (Blair et al., Reference Blair, Giraldo, Buendía, Tovar, Duque and Beebe2006a; Díaz and Blair, Reference Díaz and Blair2006; Benchimol et al., Reference Benchimol, De Campos, Morais, Colombo, Chioratto, Fernandes, Lima and Pereira2007; Zhang et al., Reference Zhang, Blair and Wang2008). Thus, it is reasonable to suggest that the bean populations studied here contain a high amount of genetic variation. Additionally, the level of genetic variation detected for five markers (BM156, BM154, BM184, PVag001 and BM114) exceeds that reported for these markers in previous literature (Yu et al., Reference Yu, Park, Poysa and Geps2000; Gaitán et al., Reference Gaitán, Duque, Edwards and Tohme2002).
On the other hand, the locus PVag004, presumably associated with the arcelin, phytohaemagglutinin and α-amylase inhibitor gene family, exhibited a multi-banding pattern and, consequently, was excluded from the analyses. However, its alleles 184, 195 and 207 bp, presumably associated with resistance to bruchids (Zabrotes subfasciatus (Boheman) and Acanthoscelides obtectus (Say)) and previously found in advanced bean lines of a Mesoamerican origin (Blair et al., Reference Blair, Prieto, Diaz, Buendia and Cardona2010a), were not discovered in this study. This finding supports the general idea that there is no resistance for bruchids in most of the landraces and old cultivars produced in Nicaragua. Alleles for bruchid resistance may remain in wild relatives, which have not been subjected to domestication processes.
Populations PV0006, PV0013, PV0020, PV0021, PV0023, PV0024, PV0027, PV0029 and PV0031, in addition to having a high level of allelic variation, are classified in national and international markets as ‘rojo chile’ or ‘rojo nacional’ because of their good culinary quality and light red seed coloration of some of these populations (similar to colours 2.5R4/10, 5R3/8 and 5R3/10 in the Munsell colour charts for plant tissue, 1977). Thus, there are sufficient arguments to suggest that at least 20% of the landraces presented here are promising sources of variation and also have a high market potential.
There was a good correspondence between the genetic clustering pattern displayed in the UPGMA tree and the grouping by structure analysis, even though they use different genetic parameters to infer population structures. Structure analysis is a Bayesian model-based method that uses genotypes from unlinked markers, demonstrating the presence of a population structure, identifying distinct genetic populations, assigning individuals to populations, and identifying migrants and admixed individuals (Pritchard et al., Reference Pritchard, Stephens and Donnelly2000). In contrast, the UPGMA tree was inferred based on pairwise genetic distances among populations (Tamura et al., Reference Tamura, Peterson, Peterson, Stecher, Nei and Kumar2011). Nonetheless, both approaches brought similar results for genetic clustering. Only six out of 42 populations mismatched these two clustering proposals (PV0014, PV0016, PV0021, PV0022, PV0027 and PV0035).
The number of clusters detected in the structure analysis was less than expected (K = 3). Basically, common bean populations were structured into two main groups. AMOVA analysis omitting populations from cluster 1 showed that the within-population variation is 47.7%, i.e. about half of the variation represents the variation among populations within clusters. The geographic origin of these landraces was a criterion for selection in this study. We tried to cover most Nicaraguan regions where bean production takes place. However, the region of origin does not influence the genetic structure, as populations from different agroecological regions can be found in the same cluster. On the contrary, other studies have found a good congruence between the genetic structure and geographic location (Santalla et al., Reference Santalla, De Ron and De La Fuente2010). In addition, it was surprising that concerning seed weights, eight out the ten populations with the biggest seeds were plotted into cluster 2 and the ten populations with the smallest seeds were plotted into cluster 3. This tendency proposes that seed weights are connected with the population structure. Such marker–trait relationship agrees with other studies that have found that seed weight is the main factor influencing the genetic structure (Díaz and Blair, Reference Díaz and Blair2006; Santalla et al., Reference Santalla, De Ron and De La Fuente2010).
Farmers who perform on-farm conservation modify the genetic structure of landraces by selection in response to their preferences and interests (Negri and Tiranti, Reference Negri and Tiranti2010). Similarly, when African landraces (introduced from Mesoamerican and Andean centres) were analysed, there were clues to ascribe many changes in genetic structure to farming conditions and preferences for specific types of beans, expressed in seed sizes and colours (Asfaw et al., Reference Asfaw, Blair and Almekinders2009; Blair et al., Reference Blair, Gonzales, Kimani and Butare2010b). Under the perspective that seed weight is a trait related to quality of sowing material and particular food preferences, it is a trait probably affected strongly by Nicaraguan farmers' selection.
Another interesting observation was obtained from marker BM114. This marker possessed 13 different alleles in total, the same number found by Blair et al. (Reference Blair, Giraldo, Buendía, Tovar, Duque and Beebe2006a) when working on almost the same number of populations (43). The allele 248 bp had a frequency above 0.8 exclusively in the populations with a known shorter period of time to start flowering. These populations are quite popular among Nicaraguan farmers and they are called ‘frijoles cuarenteños’, the name suggesting that the populations start to flower earlier than improved cultivars. Precisely, this marker was also previously associated with QTLs for days to flowering in populations mapped by Blair et al. (Reference Blair, Iriarte and Beebe2006b).
Finally, the structure inference and the genetic parameters estimated from the studied populations suggest that most genetic diversity described for common bean landraces could be more efficiently captured by selecting a few numbers of populations in equal numbers from both clusters using an appropriate number of individuals. MRA showed that genotypes AG1 138 , BM156 224 , BM156 222 , PVag001 278 , BM205 131 and BM205 133 are correlated with seed features (seed length, width and weight). The first two traits influence the final seed weight. As an example, when five populations with biggest seeds and five populations with smallest seeds are plotted, it is noticeable that the model based on genotype BM205 131 explains a good proportion of the observed variation (Fig. 2). An increase in the frequency of this genotype predicts gains in seed weights. Markers AG1, BM205 and BM156 have been previously mapped in Mesoamerican populations and highly significant associations with QTLs for seed weight have been found (P < 0.001; Blair et al., Reference Blair, Iriarte and Beebe2006b, Reference Blair, Diaz, Buendia and Duque2009). In the same way, marker BM156 has been reported to be associated with high grain productivity when identifying QTLs for high yield in beans of the type Carioca (Rodrigues et al., Reference Rodrigues, Dos Santos, Ramalho, Amorim and Silva2007; Pereira et al., Reference Pereira, Dos Santos, De Souza and Lima2008; Torga et al., Reference Torga, Dos Santos, Pereira, Furtado and Leite2010).
Similar approaches to reveal germplasm-regression-combined marker–trait associations as used here have been reported with successful results in many other species and traits, such as rice (Oryza sativa L.), alfalfa (Medicago sativa), oat (Avena sativa L.) and sea buckthorn (Hippophae L.) (Virk et al., Reference Virk, Ford-Lloyd, Jackson, Pooni, Clemeno and Newbury1996; Maureira-Butler et al., Reference Maureira-Butler, Udall and Osborn2007; Achleitner et al., Reference Achleitner, Tinker, Zechner and Buerstmayr2008; Ruan et al., Reference Ruan, Li and Mopper2009). The validation of the linkage between important traits and molecular markers is crucial not only for plant breeding proposes, but also for the characterization of potential germplasm. The high correlations found in this study in common beans suggest that four of the markers examined here could be important for marker-assisted selection for seed weight in segregating populations derived from our collection. Of course, it is necessary to test a higher number of markers and to include additional quantitative traits (pods per plant, seeds per pod, seeds per plant, for instance) in order to have a better estimation of breeding potential. Yet, an analysis of the available trait and marker information could be advantageous before performing extensive phenotypic testing. Seed weight is an important yield-determinant component in common beans (White and Izquierdo, Reference White, Izquierdo, van Schoonhoven and Voysest1991; Dalla-Corte et al., Reference Dalla-Corte, Moda-Cirino, Arias, De Toledo and Destro2010). Even though there are many factors influencing the yield, a high diversity in seed weight is an important indicator to consider when selecting potential genetic variation.
In conclusion, there is promising genetic diversity in the common bean collection described in this study. This genetic variation is higher than that reported in most previous studies, which highlights the importance to conserve these materials. A special attention should be paid to those diverse populations that are attractive in national and international markets and are present in both discovered clusters. This diversity could be maximized in breeding programmes as long as this molecular information is included as criteria for germplasm selection. The significant correlation of four microsatellite markers with seed features suggests that marker-assisted selection for these quantitative traits could be straightforward. Yet, additional marker–trait associations with other yield components should be addressed in order to enforce this practice.
Acknowledgements
We would like to thank the authorities from the Nicaragua–Finland Agrobiotechnology Program (NIFAPRO), the Finnish Ministry of Foreign Affairs, the University of Helsinki and the Embassy of Finland in Nicaragua for supporting this study. We are grateful to the National Agrarian University, Nicaragua for collaborating with us during germplasm collection. Finally, we extend our gratitude to bean farmers who kindly contributed small amounts of seeds for this study.
