The Spanish core collection of beans includes 202 accessions belonging to 51 market classes (Santalla et al., 2001) from all areas of Spain where beans are cultivated. About 75% are of Andean origin and 25% are of Mesoamerican origin (Pérez-Vega et al., 2009).

The 202 accessions were sown in Sabadell (Northeast Spain: 41° 32' 50.7" N, 2° 4' 14.7" E), a location with loam soil with abundant Ca, low P, and a mild Mediterranean climate that allows both short- and long-cycle materials to develop to maturity. We used a randomized two-block design with 21 effective plants per plot. The experiment was conducted at a low density (29,167 plants/ha) to facilitate individualized recording of data. A vertical plastic net was used to trellis the accessions with indeterminate growth. Plants were cultivated with the traditional management in the area, including drip irrigation and fertilizer when necessary. During cultivation, growth habit (Singh, 1982) was recorded.

Seeds were harvested by plots, but owing to the lack of significance of the block factor for agronomic traits, the seeds from the two blocks for each accession were combined to ensure a sufficient sample for the culinary and sensory analyses.

To estimate the variation, we calculated the means, variances, and covariances of the recorded traits.

Accessions were grouped successively in function of the type of phaseolin (indicating Andean or Mesoamerican origin of domestication), seed-coat color, and growth habit. For each type of classification, analysis of variance (ANOVA) was done based on the linear model y_ij=µ + α_i+ ε_ij, where yi was an individual level for a specific trait, µ was the grand mean, α_i was the effect of i group based on type of phaseolin, seed-coat color, or growth habit, and ε_ij was the random error for i groups with j replications of the model following a N ~ (0, σ²). All factors were considered fixed. Finally, for the type of phaseolin and seed-coat color, we performed normalized principal component analysis (PCA) and we calculated 95% confidence ellipses around each cluster of accessions. We used the R software (R Development Core Team, 2007) Agricolae, pcaMethods, and Ellipse packages for all statistical analyses (Murdoch & Chow, 2007; Stacklies et al., 2007; Mendiburu, 2010).

At the end of the trial, complete datasets were available for 174 of the 202 accessions in the core collection. For visual traits of the raw seeds, the coefficients of variation (CVs) ranged from 11% for width to 162% for the percentage of white color (Table 1). For the culinary traits, the CVs ranged from 13% for water absorption at 24 h to 81% for water absorption at 3 h (Table 1). For the sensory traits after cooking, the CVs ranged from 24% for aroma to 76% for seed-coat roughness (Table 1).

Having data about a large number of accessions with high variation in all traits enabled us to perform a robust study of the correlations among the sensory and culinary traits. Although r values ≥ 0.25 were significant different from zero at p<0.001, for practical purposes, we discuss only absolute rvalues ≥ 0.50 (Table 2).

The beans with the greatest proportion of white surface were also the brightest (r=0.60); these beans were also less mealy (r=-0.60) than darker beans (Table 2). The beans with rougher seed coats and faster water absorption were also less mealy (r=-0.60, r=-0.53). The size of the seeds (100-seed weight) correlated positively with seed-coat perceptibility (r=0.67). Water absorption during cooking correlated positively with cooking time (r=0.57) and negatively with the percentage of whole beans after cooking (r=-0.54). Flavor correlated negatively with seed-coat roughness (r=-0.55) and positively with aroma, seed-coat perceptibility, and mealiness (r=0.71, r=0.57, r=0.54).

Our study comprises 131 accessions domesticated in the Andes and 43 domesticated in Mesoamerica (Pérez-Vega et al., 2009). The results showed significant differences between the two origins for the visual traits of the raw seeds (p<0.05 for length/width and p<0.001 for the others), except seed curvature and percentage of white surface (Table 3). The PCA also showed that Mesoamerican materials tended to have smaller seeds (Fig. 1a).

With regard to the sensory traits of the cooked beans (Table 3), we found that materials of Andean origin had lower seed-coat brightness (p<0.01) and seed-coat roughness and higher seed-coat perceptibility, mealiness, flavor, and aroma (all significant at p<0.001) than materials of Mesoamerican origin. These results are reflected in the PCA done on the sensory traits of the cooked beans and phaseolin type (Fig. 1b). No significant differences were found in any culinary trait, and the PCA failed to reveal any clear clustering among materials of Mesoamerican or Andean origin based solely on culinary traits.

The ANOVA on the means of the sensory and culinary traits of the accessions distributed into 13 color categories of the raw seeds revealed few significant differences between groups for most traits, although the rate of water absorption was generally lower (40-62, p<0.001) in brown, vinous brown, gray-brownish to greenish, purple, and black seeds and higher (84-93, p<0.001) in yellow and white seeds. Nevertheless, some black beans showed a high rate of water absorption.

In the PCA performed on sensory traits of the cooked seeds in function of the color classification of the accessions, white beans tended to have rougher and brighter seed coats and to be less mealy than more pigmented beans, while pigmented beans tended to have a stronger aroma and flavor than white beans (Fig. 2a). On the other hand, the PCA of the culinary traits (Fig. 2b) also confirmed that white beans tended to have higher rates of water absorption and to yield a lower percentage of whole beans after cooking (Fig. 2b).

Our study includes 29 accessions with determinate growth (Type I) (Singh, 1982) and 111 accessions with indeterminate growth (Type IV). The ANOVA to compare the mean values of the sensory and culinary traits between materials with determinate and indeterminate growth found significant differences only in seed-coat perceptibility (p<0.001) and flavor (p<0.01), but these differences were small (for seed-coat perceptibility: 3.7 in indeterminate growth vs. 2.8 in determinate growth; for flavor: 6.6 in indeterminate growth vs. 6.1 in determinate growth).

The Spanish core collection of beans, derived from screening more than 3000 accessions fundamentally through the criteria of collection area and appearance of the raw seeds and later fine-tuned with information from molecular markers of genetic background, also has wide variability in sensory and culinary traits. The especially high variation in the traits recorded in the cooked beans suggests that the cooking process generates variation in the characteristics of the seed that is added to the variation that was present before cooking.

The traits 100-seed weight and seed width, length, and thickness were included in the selection criteria for the core collection. The CVs for these traits were among the lowest for all the traits studied, coinciding with the findings reported by Casquero et al. (2006). The high CVs in culinary and sensory traits of the cooked seeds support the idea that the current core collection encompasses a large proportion of the variation for these traits that were not initially included in the selection criteria.

For the trait water absorption, the CVs were lower at 24 h than at 3 h, 6 h, or 9 h. This suggests that differences between accessions for this trait are more substantial in the early part of the soaking process. The rate of water absorption is a variable that we wanted to explore as a synthetic value for the differences found at different times in the process. To this end, we analyzed the correlations between this variable and two absorption constants proposed by Peleg (1988): K₁ (a constant related to mass transfer rate) and K₂ (a constant related to maximum water absorption capacity). The rate of water absorption correlated negatively with K₁ (r=-0.81, p<0.001) and positively with K₂ (r=0.80, p<0.001), showing that the rate of water absorption is a good synthetic trait that summarizes the relevant information.

In the overall set of the core collection, we found a wide variety of combinations between visual traits of the raw seeds, sensory traits of the cooked seeds and culinary traits, although some combinations were more common than others. The strongest correlations were found between seed-coat brightness and mealiness (r=-0.73, p<0.001) and between flavor and aroma (r=0.71, p<0.001). Plans et al. (2014) reported similar results for flavor and aroma (r=0.62, p<0.001) and also reported that flavor correlated with seed-coat roughness (r=- 0.4, p<0.01) and mealiness (r= 0.68, p<0.001). In our study, flavor also correlated with seed-coat roughness (r=-0.55, p<0.001) and with mealiness (r=0.54, p<0.001). We found that 100-seed weight correlated with seed-coat perception (r=0.67, p<0.001), % of white correlated with mealiness (r=-0.60, p<0.001), and with seed-coat brightness (r=0.60, p<0.001). Mealiness also correlated negatively with seed-coat roughness (r=-0.60, p<0.001), as previously reported by Rivera et al. (2013) in an earlier study done in Catalan landraces.

The relation between seed size and rate of water absorption is controversial (Sefa-Dedeh & Stanley, 1979; Del Valle et al., 1992; Abdel Kader, 1995; Berrios et al., 1999). We found a very low correlation between the rate of water absorption and seed volume (r=-0.19, p<0.05), suggesting that factors other than surface area and volume are also involved; these factors might include the shape of the hilum and micropyle or the structure and composition of the seed coat and cotyledon and their thickness (Sefa-Dedeh & Stanley, 1979).

Phaseolus vulgaris L. was domesticated independently in the Andean and Mesoamerican regions, resulting in two highly differentiated gene pools that can be distinguished by the phaseolin content (Gepts & Blis, 1986; Gepts et al., 1986; Koenig et al., 1990). Using this approach, these authors found that seeds of Mesoamerican origin tend to be smaller than those of Andean origin. We found significant differences between the two origins for all traits related to size and shape (p<0.05 for length/width and p<0.001 for the others), except seed curvature, a trait not included in previous studies and percentage of white in the seed surface, although the mean values for this trait differed greatly between Andean and Mesoamerican materials, these differences were not significant (Table 3).

Our results showed significant differences between the two origins for all the sensory traits of the cooked beans, but none for the culinary traits. Materials of Andean origin had lower seed-coat brightness (p<0.01) and seed-coat roughness, and greater seed-coat perceptibility, mealiness, flavor, and aroma (all significant at p<0.001). These differences had not been reported before and add new criteria for the description of the two groups.

The color of beans is attributed to the presence and quantity of polyphenols, such as condensed tannins (Espinosa-Alonso et al., 2006), which reside mainly in the seed coat and are found in greater concentrations in black, red, and bronze seeds (Bressani & Elias, 1980). Seed-coat color has been related with other traits of interest like Pythium ultimum resistance (Lucas & Griffiths, 2004; Campa et al., 2010) or seed composition (Casañas et al., 2013; Hacisalihoglu & Settles, 2013).

We found few significant differences in culinary and/or sensory properties in relation to color. Beans with a high proportion of white surface area tended to be brighter, rougher, and less mealy; they also absorbed water faster and yielded a low percentage of whole beans after cooking (Table 2; Figs. 2a and 2b). Stanley (1992) attributed the lower rate of water absorption in colored seeds to the formation of macromolecules from tannins and other molecules such as proteins and pectins. These macromolecules strengthen the cell wall and reduce the absorption of water, impeding the separation of cells and softening. However, Espinosa-Alonso et al. (2006) reported that the variation in polyphenolic contents found in their studies was more related to the rest of the genetic background than to color. In a previous study (Rivera et al., 2013), our group found a similar correlation between the percentage of white in raw seeds and the percentage of whole beans after cooking (r=-0.47, p<0.05) to those found in the present study.

Growth habit did not correlate with culinary or sensory traits. These results do not explain the predominance of beans with indeterminate growth among the most prestigious varieties in Spain (PGI Judías de El Barco de Ávila, PGI Alubia de La Bañeza-León, PGI Faba Asturiana, PGI Faba de Lourenzá, and PDO Mongeta del Ganxet) (Casquero, 2012). This suggests that we should probably refer to agronomical, historical, or cultural reasons to explain this point.

Our results add information on the Spanish core collection of beans, increasing its value as a source of material that can be used to obtain new varieties. Breeding programs benefit greatly from more complete characterization of the accessions, including information about culinary aspects, the appearance of the raw seed, texture, and flavor of the cooked seed or resistance to disease. Moreover, crossing the materials in the collection will help identify genes related to sensory and culinary quality. The recent sequencing of the genome of Phaseolus vulgaris L. (Schmutz et al., 2014) and the identification of quantitative trait loci related to sensory and culinary quality (Posa-Macalincag et al., 2002; Garcia et al., 2011; Casañas et al., 2013) should pave the way for the identification of genes and the development of internal markers that will allow them to be managed easily.

The traits detailed in the accessions of the Spanish core collection can be consulted on the Spanish National Plant Genetic Resource Center’s website www.crf.inia.es/crfesp/paginaprincipaljudia.asp.