Multivariate Characterization of Bean Varieties According to Yield Production , Mineral and Phenolic Contents

Nutrientes minerais (Fe, Ca, Cu, Mn, Mg e Zn) e compostos fenólicos (fenóis totais e taninos) de dezesseis variedades de feijões, cultivados sob condições controladas, foram quantificados. Os dados obtidos das determinações químicas e das produtividades dos feijões foram avaliados empregando métodos multivariados: análise de componentes principais, análise de agrupamentos e análise de correlação canônica. Três grupos de cultivares foram reconhecidos e tanto os minerais Fe, Zn, Cu e Mn quanto os rendimentos dos feijões foram os principais responsáveis por esta discriminação. Variedades promissoras de feijões para programas de biofortificação ou uso direto em alimentos fortificados foram identificadas principalmente no grupo I, no qual os cultivares Aporé (Fe e Mg), Bambuí (Zn) e Valente (Fe e Cu) conciliaram boa produtividade e altos níveis de minerais. O uso da análise de correlação canônica possibilitou detectar relações significativas entre produtividade dos feijões, taninos e os metais Mg, Zn e Mn.


Introduction
The common bean (Phaseolus vulgaris L.) is one of the most important sources of proteins, calories, B-complex vitamins, and minerals in Latin America.In Brazil, the bean consumption per capita is about 17 kg/year, 1 which makes this legume essential against severe iron deficiency anaemia, a critical public health problem in South and Central America, affecting as many as 25% of pregnant women and 40% of children under 5 years old. 2 In developing countries, beans are the best non-meat source of iron, providing 23-30% of daily recommended levels in a single serving. 1 Common beans also contribute to human nutrition with other important minerals, such as Zn, Ca, Mg, Mn and Cu. 3 Mineral contents in bean grains can vary largely depending on the varieties, 4,5 and on environmental factors, such as soil composition. 6Taking into account the nutritional value of beans, another important aspect needs to be considered: the presence of anti-nutritional components, such as tannins, phenols, phytic acids, and calcium, which depress Fe bioavailability. 7he majority of bean breeding programs developed in Brazil have focused on improving resistance to the disease and cooking performance, in addition to increasing crop yields and maturation rate. 8However, the nutritive quality of the seeds has recently been considered in relation to varietal selection. 1This process is named biofortification, which increases the amounts of bioavailable essential nutrients through soil fertilization or genetic selection. 9,10he objective of this study was to quantify the content of some mineral nutrients (Fe, Zn, Mg, Mn, Ca and Cu) and compounds with antinutritional value (total phenols and tannins) of sixteen bean cultivars grown in Brazil and their respective yield production.Chemometric tools (principal component and cluster analyses) were applied to detect pattern distributions of bean cultivars and to identify which parameters differentiate the natural groups.In addition, the canonical correlation analysis was used to study the influence of yield seed production and phenol contents on mineral nutrients.
The multivariate methods of analysis can be a useful tool in selecting cultivars which conciliate good nutritional value and high yields for direct use as biofortified food or potential varieties for biofortification breeding programs.

Plant materials
The experiment was carried out in Embrapa Arroz e Feijão, Goiás State.Bean plants of sixteen cultivars (Pérola, Aporé, Rudá, Valente, Diamante Negro, Xamêgo, Bambuí, Jalo Precoce, BRS Executivo, SIN 15, AFR 245, WAF 69, WAF 75, A 195, AND 676, and DRK 18) were raised from seeds in plastic pots (diameter 30 cm, height 35 cm) with three seedlings per pot and three pots for each variety.The local soil was enriched with a standard solid fertilizer mixture N:P 2 O 5 :K 2 O, ratio 10:11:5, and plants were watered once a day.The experiment was performed in a greenhouse with natural photoperiod and light intensities.The pods were harvested from 75 to 120 days, when seeds were still fresh.They were subsequently dried at room temperature in the shade for 7 days.

Extraction of phenols
Powdered and dried whole bean seeds (2 g) were extracted at room temperature with 80% v/v aqueous methanol in an ultrasonic bath.Samples were extracted with 10 mL of solvent for 1 h.The extract was separated from the solid residue by centrifuging at 4000 rpm for 15 min and transferred to a 25 mL volumetric flask.The same procedure was repeated twice with 10 mL of solvent for 30 min and with 5 mL for 30 min.The extracts were combined to a final volume of 25 mL. 11Immediately after the extraction the samples were analyzed for total phenols and tannins.

Phenol analysis
Total phenols were quantified by the Folin-Ciocalteu method.A mixture of 0.5 mL of each extract, 0.5 mL of Folin-Ciocalteu reagent, and 10 mL of Na 2 CO 3 1 mol L -1 was placed in a 25 mL volumetric flask.Once the solutions had reacted for 1 h, they were measured at 750 nm via a Beckman DU-70 spectrophotometer. 11Tannins were measured by the protein precipitation assay with the use of BSA. 12The extract (2.0 mL) was mixed with 2.0 mL of the BSA solution (1.0 mg mL -1 ) and was left to stand for 15 min.The precipitate was separated by centrifuging for 15 min at 4000 rpm and later dissolved in 4.0 mL of sodium dodecyl sulphate solution.The absorbance was measured at 510 nm after 15 min of the addition of the FeCl 3 reagent.All samples in the two assays were analyzed in triplicates and standard curves were constructed with tannic acid.The results were expressed in mg of tannic acid equivalent g -1 dry seed (mg g -1 ).

Mineral analysis
Bean seeds were rinsed four times with deionised water, dried in a ventilated oven at 60 o C for 24 h, and ground in an agate mortar with an agate pestle to pass an 80 mesh sieve.Powdered samples (1.0 g) were hotplate wet-digested in an open-vessel at 110 o C for 5 h, using nitric acid-hydrogen peroxide 5:3 v/v.The acid digests were analyzed for Fe, Zn, Ca, Cu, Mg and Mn by flame atomic absorption spectrometry (Instrumentos Científicos CG AA 7000BC, Brazil).In the case of calcium, lanthanum chloride was added to the mineral solution (final concentration 1% m/v) to avoid interference from phosphate.Calibration of measurements was performed using commercial standards.All measurements were carried out using standard flame operating conditions, as recommended by the manufacturer.Vol.21, No. 10, 2010

Statistical analysis
Principal component analysis (PCA) and cluster analysis (CA) were applied to examine the interrelationships between experiment yields, mineral nutrients, and phenol contents using the software package SPAD.N version 2.5, CISIA, France. 13Nearest neighbour complete linkage technique by Benzécri algorithm was used as an index of similarity. 14Hierarchical clustering was performed according to Ward's variance minimizing method. 15elationships between phenol levels, mineral nutrients, and crop yields were obtained via canonical correlation analysis (CCA) using the SAS CANCORR procedure.The magnitude of structure correlation coefficients (canonical loadings) was used to explain canonical variates.The predictive ability was assessed by canonical redundancy analysis with standardized variance coefficients.Prior to the multivariate analysis, the data was preprocessed by means of auto-scaling and mean centering.
Average multiple comparisons were established by univariate analysis of variance (ANOVA) using SAS GLM analyses.All data were checked for homoscedasticity with Hartley's test (Table 1) or Cochran's test (Tables S1 and  S2).These tests revealed significant departures from the basic assumption for the variables total phenols, Fe, Zn, Mn and Mg (Table 1), which were transformed by rank (TP), reciprocal (Fe, Zn and Mn), and reciprocal square root (Mg).Whenever a difference was established, a post-hoc Tukey test was performed.Results are shown as mean values and are joined by the standard deviation of independent measurements in some cases.P-values below 0.05 were regarded as significant.

Results and Discussion
The public health problem of iron deficiency in Brazil has been controlled by two main strategies: the distribution of ferrous sulphate supplements and the fortification of wheat and corn flour with iron and folic acid. 16Despite some successful results, these programs require continuous investment and solid governmental support for long periods.In addition, they focus solely on iron deficiency anemia, whereas other metal deficiencies such as Zn, Cu and Ca have not received the same attention.An alternative for the prevention and treatment of these deficiencies is the biofortification of edible foods that are highly consumed by the population, which is the case of beans in Brazil, with essential minerals. 1,10n order to evaluate potential bean cultivars for biofortification programs, sixteen bean varieties were chosen and their mineral nutrients and phenol contents were quantified and analysed together with yield grain production via chemometric methods.Results obtained from the cultivation experiment are listed in Table S1 (see supplementary information, SI) and the concentrations of six metals (Fe, Zn, Ca, Mg, Mn, and Cu) as well as the levels of total phenols and tannins are summarized in Table S2 (SI).

Chemometric analysis
The results obtained from PCA and cluster analysis via Ward's technique (16 samples ×14 variables = 224 data) have revealed the existence of a high chemical variability throughout the cultivars of data collection.Figure 1 shows the relative position of the sampling bean cultivars in the discriminant space in relation to a diaxial system originated in the PCA.First PC accounts for 38.1% of total variance and discriminates well above the 99% confidence level cultivars with high production of pods and seeds from varieties with lower yields.In addition, the second PC accounts for 20.3% of total variance and showed a gradient of mineral nutrient concentration, in which cultivars rich in metals load on the PC-2.
The Pearson correlation showed two strong positive correlations, Fe with Cu (p < 0.01) and Mn with Zn (p < 0.01).The latter was also detected for Andean origin cultivars, 3 but no relationship was observed between iron and zinc, in contrast to previous studies. 3,4,10n fact, the canonical correlation analysis (Table 2), which was applied in order to study the influence of the cultivation yield and phenol contents on the amounts of mineral nutrients, revealed that the weight of pods and dried seeds have a direct correlation with magnesium and are related to cultivars from Cluster I.In addition, tannins and the weight of aerial parts from the first set and Mn and Zn from the second set load positively onto the first canonical variable, which are related to varieties from Cluster II.The data clearly indicate that the higher the weight of pods and seeds, the lower the levels of manganese and zinc; similar trends were detected for Zn and Cu in relation to seed yields. 17Furthermore, there was no correlation between tannins and iron, which is in agreement with a previous study. 4

Relationships between bean cultivar productivity and nutritional value
The use of chemometric methods of analysis allowed the classification of the cultivars into three natural groups, which are discriminated by yield variables (number and  The challenge of biofortification programs is to produce grains with superior amounts of several essential mineral elements in high yields, even when plants grow in infertile soils. 10The increase in concentration of simultaneous metals should be followed by the grain yield or it should at least be unaffected like other qualities. 18The relationship between minerals and yield is still uncertain, Fe seems to be yield-independent, whereas Zn, Mn (Table 2), Cu, and B are negatively correlated with yield, 17 and Mg is positively correlated (Table 2).
High amounts of anti-nutrients such as tannins can decrease the bioavailability of metals and proteins from the seeds, 7 as these compounds are able to form chelates with Fe(III), Cu(II) and Zn(II) as well as insoluble complexes with dietary proteins and enzymes. 19,20The negative correlation between tannins and yield variables, weight of dry seeds and pods (Table 2), reflects the plant strategy to invest in the primary metabolism for growth and reproduction instead of in secondary compounds for defense.This trend is normally observed when plants grow in fertile soils without herbivore stress. 21

Conclusions
Six mineral elements, phenol contents, and yields were used to characterize sixteen bean cultivars by the application of a chemometric approach.Employing multivariate methods of analysis, such as PCA, CA, and CCA, it was possible to highlight correlations between the investigated parameters and to outline some patterns.The statistical evaluation led to the distinction of three groups of cultivars, in which mineral nutrients Fe, Zn, Cu, and Mn and yields were most responsible for their discrimination.These methods are useful for the detection of potential bean cultivars for biofortification programs or for direct use in fortified food and supplements which may prevent or treat mineral deficiencies.

Figure 1 .
Figure 1.PCA ordination of metal, phenol, yield variables, and bean cultivars to which cluster it belongs: I (), II (), and III (). a Axes refer to ordination scores obtained from cultivar samples.b Axes refer to ordination scores obtained for the discriminant variables which are represented as vectors from the origin.TP = Total Phenols, T = Tannins, Pn = Pods number, Pw = Pods weight, Sn = seeds number, FSw = Fresh seeds weight, DSw = Dried seeds weight.

Figure 2 .
Figure 2. Dendrogram obtained by cluster analysis of bean cultivars using Ward's linkage method.

Table 1 .
Total phenols and tannins (mg tannic acid g -1 seed, dry wt), mineral elements (mg 100 g -1 ), and cultivation yields in the clustered P. vulgaris seeds a a Based on original data.Means followed by the same letter in the rows did not share significant differences at 5% probability by Tukey's test.

Table 2 .
Canonical structure (loadings) of phenolic and yield discriminants and mineral nutrients with their canonical variates.