Variation in Pea (Pisum sativum L.) Seed Quality Traits Defined by Physicochemical Functional Properties

Pea is one of the most produced and consumed pulse crops around the world. The study of genetic variability within pea germplasm is an important tool to identify outstanding accessions with optimal functional and nutritional qualities. In the present study, a collection of 105 pea accessions was analysed for physicochemical properties, pasting viscosity, and basic composition parameters. While pasting viscosities were negatively correlated to hydration capacity, cooking time, and basic composition, a positive correlation was found between the hydration capacity and the basic composition parameters. Basic composition (protein, fibre, fat, and resistant starch) parameters were further evaluated regarding seed trait morphology, namely, seed shape, colour, and surface. Allelic characterisation at the r and rb genetic loci was performed in a subgroup of 32 accessions (3 phenotyped as smooth and 29 as rough seeded), revealing that none of the initially classified rough-seeded accessions were rb mutants, 19 were r mutants, and 13 were neither r nor rb. Despite their initial phenotypic classification, the 13 accessions genetically classified as smooth behaved differently (p < 0.05) to the 19 r mutants in terms of physicochemical properties, pasting viscosity, and basic composition parameters. Using multivariate analysis of the most discriminatory parameters for the food-related traits studied, the best-performing accessions at functional and nutritional levels were identified for future plant breeding to improve field pea production and consumption.


Introduction
Field pea (Pisum sativum) is an increasingly important legume crop grown around the world, having a total harvested area of 8 million ha and a total production of 16 million tonnes per year [1]. Currently, Europe secures a pea production share of about 44%, followed by America and Asia. Of the total 5.1 million tonnes of pulses produced in Europe, field pea accounts for more than 40% of the production volume [1,2]. To increase their consumption and avail consumers of their functional and nutritional advantages, legume (pulse) seeds are being processed into flours to be used as ingredients Figure 1. Pea seed categories defined for colour, surface, and shape. Geometric representations correspond to seed shape (elliptical, cylindrical and rhomboid).

Statistical Analysis
Data were analysed using GraphPad Prism version 8.1.2 for Mac OS X (GraphPad Software, La Jolla California USA, www.graphpad. com). The variation within each seed trait was analysed by one-way analysis of variance (ANOVA). Physicochemical parameters determined for the genetically classified rough-and smooth-seeded accessions were compared by unpaired t-tests using the Holm-Šidák method. Statistical significance was considered at p < 0.05.
Overall variation of the physicochemical, cooking, rheological, and basic composition characteristics was assessed by principal component analysis (PCA) using Tanagra data mining software, version 1.4.5 (Lyon, France) [24]. A subset of lines (32) selected as having "rough" or "wrinkled" seeds (29 accessions) and three smooth (round-seeded) accessions (259, 286, 289) with low viscosity profiles were genotyped at the r and rb genetic loci to determine the nature of the mutation which was impacting on the seed surface trait. The r locus encodes starch-branching enzyme I, whereas the rb locus encodes the large subunit of ADP-glucose pyrophosphrylase [17]. Seed meals were used for the extraction of DNA for 29 samples, and leaf DNA was extracted for the remaining three lines where high-quality DNA could not be extracted. A triplex assay was used to determine the nature of the r locus, whereas the rb allele was determined by sequencing, using PCR assays as previously described [17]. Starch granule shape appearance was determined by microscopy for this set of lines to validate the genotyping results. Four pea accessions of known genotype were included as controls in the DNA assays (JI 2822, RRrbrb, simple starch grains; JI 1194, rrRbRb, compound starch grains; JI 281, RRRbRb, simple starch grains; JI 399, RRrbrb, simple starch grains).

Statistical Analysis
Data were analysed using GraphPad Prism version 8.1.2 for Mac OS X (GraphPad Software, La Jolla California USA, www.graphpad.com). The variation within each seed trait was analysed by one-way analysis of variance (ANOVA). Physicochemical parameters determined for the genetically classified rough-and smooth-seeded accessions were compared by unpaired t-tests using the Holm-Šidák method. Statistical significance was considered at p < 0.05.
Overall variation of the physicochemical, cooking, rheological, and basic composition characteristics was assessed by principal component analysis (PCA) using Tanagra data mining software, version 1.4.5 (Lyon, France) [24].

Seed and Flour Variation in Pea Germplasm
In the present work, a set of pea accessions (105) was analysed for variation in traits related to food end-use. In terms of 100-seed weight, the accessions 245, 246, 247, and 227 were the ones with lowest weight (below 10 g), while the accessions 298, 221, 216, and 217 registered weights above 28 g (Figure 2a). Low seed weight has been shown to be related to the presence of relatively small-sized starch granules per unit area [25], which might impact hydration capacity and cooking time. The hydration capacity refers to the amount of water absorbed per 100 g of whole mature (dried) seeds. The accessions 246 and 247, which showed the lowest seed weights, also displayed relatively low hydration capacity ( Figure 2b) and cooking time (Figure 2c). Despite the variability among the pea collection regarding cooking time (5-120 min), the average of 33 min is low when compared to other pulses [9]. In order to successfully substitute or partially replace wheat flour with pulse flour, it is important to understand how the resulting dough will perform in terms of its rheological properties. Pasting parameters, namely, peak, trough, breakdown, setback, and final viscosities, were analysed as these affect the processing conditions [15]. The analysis of the pasting viscosities (Table 1) showed that the peak viscosity, the maximum viscosity achieved by the samples, ranged between 83 and 4836 cP; the trough viscosity, which represents the decrease in paste viscosity caused by the disruption of starch granules, ranged between 58 and 4066 cP; the breakdown viscosity, the difference between the peak viscosity and trough viscosity, ranged between 8 and 867 cP; the final viscosity ranged from 212 to 7471 cP; and the setback viscosity, the difference between peak and final viscosity, was in the range 151-3489 cP.
The average values of the pasting properties of the 105 pea accessions were similar to those reported by others [14] and to the values obtained for other pulses, such as grass pea, chickpea, or lentil [9]. When compared to wheat varieties [26], the peak viscosity was generally lower, which could be related to differences in the amylose content of starches. In addition, the breakdown viscosity was lower for the pea collection described here (285 cP) when compared to wheat values (669 cP). This might present an advantage when incorporating pea flours into food formulations, since breakdown values are a measure of the degree of paste stability [26], and high breakdown viscosity can reduce the ability of flour to withstand heating during cooking [27]. The lower breakdown viscosities observed here are similar to those recently reported by others for pea and other pulses [28]. In order to successfully substitute or partially replace wheat flour with pulse flour, it is important to understand how the resulting dough will perform in terms of its rheological properties. Pasting parameters, namely, peak, trough, breakdown, setback, and final viscosities, were analysed as these affect the processing conditions [15]. The analysis of the pasting viscosities (Table 1) showed that the peak viscosity, the maximum viscosity achieved by the samples, ranged between 83 and 4836 cP; the trough viscosity, which represents the decrease in paste viscosity caused by the disruption of starch granules, ranged between 58 and 4066 cP; the breakdown viscosity, the difference between the peak viscosity and trough viscosity, ranged between 8 and 867 cP; the final viscosity ranged from 212 to 7471 cP; and the setback viscosity, the difference between peak and final viscosity, was in the range 151-3489 cP. The average values of the pasting properties of the 105 pea accessions were similar to those reported by others [14] and to the values obtained for other pulses, such as grass pea, chickpea, or lentil [9]. When compared to wheat varieties [26], the peak viscosity was generally lower, which could be related to differences in the amylose content of starches. In addition, the breakdown viscosity was lower for the pea collection described here (285 cP) when compared to wheat values (669 cP). This might present an advantage when incorporating pea flours into food formulations, since breakdown values are a measure of the degree of paste stability [26], and high breakdown viscosity can reduce the ability of flour to withstand heating during cooking [27]. The lower breakdown viscosities observed here are similar to those recently reported by others for pea and other pulses [28].
On average, the pea accessions contained 22% protein, 7% fibre, and 2% fat-values which are generally comparable to the values reported by others for this pulse [6,12,15,29]. The maximum resistant starch content was 7% in the pea accessions 220, 221, and 228, which also showed higher peak and setback viscosities. Also, the average for the 105 accessions was 3%, which is consistent with the observation that this pulse, when compared to others such as chickpea or lentil, has higher resistant starch percentages [6]. Once again, this is important since higher resistant starch contents lead to slower rates of digestion, enabling the use of pea starches in dietetic foods [30,31].

Correlation Analysis of Pasting, Physicochemical, and Basic Composition Parameters
Correlation coefficients estimated on the means of data from all pea accessions for pasting (trough viscosity, break viscosity, final viscosity, and setback viscosity), physicochemical parameters (100-seed weight, hydration capacity, unhydrated seeds, and cooking time) and basic composition (protein, fibre, fat, and resistant starch contents) are presented in Table 2. Table 2. Correlation coefficients of the traits † presented in Table 1 for the 105 pea accessions analysed. As previously reported [9], all viscosity parameters were positively correlated with each other (p < 0.01). These were all positively correlated to 100-seed weight (p < 0.05) but showed a significant negative correlation to hydration capacity, as well as cooking time (Table 2). This negative correlation may be due to the fact that viscosity parameters are highly related to the firmness and cooking quality of pulses [32], which is in turn also influenced by the starch composition. Here, the peak, trough, final, and setback viscosities were negatively correlated to resistant starch content (p < 0.05).
Further correlations were found when looking at the basic composition. Firstly, all nutritional components were negatively correlated to the viscosity parameters (Table 2). Concordantly, in landraces of Phaseolus bean, fat content was shown to be negatively correlated to peak viscosity, but protein content displayed a positive correlation [33]. Additional positive correlations between protein, fibre, fat, resistant starch, and hydration capacity were found (p < 0.01), in agreement with the literature [32,34]. Fibre and fat were also positively correlated with cooking time, possibly indicating the major role of these constituents in pea processing.

Seed Trait Variation
The means of the peak, trough, break, final, and setback viscosity, 100-seed weight, hydration capacity, unhydrated seeds, and cooking time for each class of seed shape, surface, and colour are presented in Table 3. Table 3. Mean peak (cP *), trough (cP), break (cP), final (FV, cP), and setback viscosities (cP), seed weight (SW, g), hydration capacity (HC, %), unhydrated seeds (US, %), and cooking time (CT, min) traits determined for 105 pea accessions within the classes distinguished by seed shape, colour, and surface † . Regarding seed shape, most accessions had ellipsoid (n = 42) and cylindrical (n = 24) shapes, and these were the types of seeds with higher viscosity values (Table 3). Irregular seeds displayed significantly longer cooking time when compared to ellipsoid or cylindrical seeds (Table 3). Irregular seeds showed higher protein and resistant starch contents when compared to ellipsoid and cylindrical seeds (Table 4). Regarding seed colour, the variability was not as high for all parameters as it was for seed shape (Table 3). However, most seeds showed a yellow green (n = 24) or green (n = 23) colour. Significant differences were found between these two groups for peak, trough, final, and setback viscosities, where yellow green seeds had higher values. They also differed in hydration capacity, where yellow green seeds had lower values ( Table 3). Analysis of the basic composition variability (Table 4) showed that dark green seeds had higher protein content than cream yellow, and light green seeds had the highest resistant starch content, this difference being significant when compared to brown-coloured seeds.

Peak
Variation in seed surface type among the lines was apparent, and the majority of seeds were smooth (n = 64). These seeds had higher viscosity parameters (p < 0.05) and shorter cooking time (p > 0.05) ( Table 3). When looking at basic composition, the only significant difference detected between the two seed surface types was that smooth seeds have a lower protein content when compared to rough seeds (Table 4).

Characterisation of Allelic Variation at the r and rb Genetic Loci
For the 32 selected lines, genotyping analysis revealed that none of these were rb mutants, 19 were r mutants, and 13 were neither r nor rb ( Table 5). The seed granule morphology scores confirmed the genotyping results, where the r accessions showed a compound granule structure, and those lines which were neither r nor rb showed a simple starch granule structure ( Table 5). The controls included three rough-seeded lines (JI 1194, JI 2822, and JI 399 as one r mutant and two rb mutant lines) and one wild-type smooth-seeded line (JI 281). On the basis of these results, it seems likely that 13 of the 29 classified as having "rough-seeded" phenotypes were genetically round (smooth) seeded and that the three lines classified as having "smooth-seeded" phenotypes (259, 286, 289) with low viscosity profiles were genetically "rough-seeded".
The 13 lines initially classified as "rough"-seeded which were scored as RRRbRb, genetically classified as "smooth", demonstrated the strong environmental effect on the seed surface trait phenotype which led to difficulties in obtaining consistent classification scores across different growth seasons or generations. It has been shown previously that two mutations (r, rb) account for the round-/wrinkled-seeded phenotype in a pea germplasm resource [17]. Premature harvest of round-seeded genotypes or premature desiccation as a result of stress will lead to a wrinkled appearance of the seeds when residual water is lost from seeds more rapidly than would be the norm.
These 13 genetically smooth-seeded accessions revealed different behaviour in regards to all of the components analysed in the present study when compared to the 19 rough-seeded accessions (Figure 3). For example, when looking at the physicochemical parameters (Figure 3a), smooth seeds had significantly lower hydration capacity, higher unhydrated seed percentage, and shorter cooking time. Also, in the viscosity profiles (Figure 3b), the smooth-seeded accessions exhibited significantly higher values (with the exception of accession 247). Finally, significant differences were also found between these two groups in basic composition (Figure 3c). Smooth seeds had significantly lower levels of fibre, fat, and resistant starch (but not protein) when compared to the rough-seeded accessions. It is interesting to note that three seed samples (226, 247, 299) posed a problem for DNA preparation, and leaf DNA was required to enable the genotyping assays (Table 5). This may reflect differences, either genetic or environmental, in the nature of starch in the seeds of these lines, which can interfere with the isolation and purification of other seed components. appearance of the seeds when residual water is lost from seeds more rapidly than would be the norm. These 13 genetically smooth-seeded accessions revealed different behaviour in regards to all of the components analysed in the present study when compared to the 19 rough-seeded accessions ( Figure 3). For example, when looking at the physicochemical parameters (Figure 3a), smooth seeds had significantly lower hydration capacity, higher unhydrated seed percentage, and shorter cooking time. Also, in the viscosity profiles (Figure 3b), the smooth-seeded accessions exhibited significantly higher values (with the exception of accession 247). Finally, significant differences were also found between these two groups in basic composition (Figure 3c). Smooth seeds had significantly lower levels of fibre, fat, and resistant starch (but not protein) when compared to the rough-seeded accessions. It is interesting to note that three seed samples (226, 247, 299) posed a problem for DNA preparation, and leaf DNA was required to enable the genotyping assays (Table 5). This may reflect differences, either genetic or environmental, in the nature of starch in the seeds of these lines, which can interfere with the isolation and purification of other seed components. Bars are means ± SD. * and *** indicate significant differences at p < 0.05 and p < 0.001, respectively, by unpaired t-tests using the Holm-Šidák method. Bars are means ± SD. * and *** indicate significant differences at p < 0.05 and p < 0.001, respectively, by unpaired t-tests using the Holm-Šidák method.
The molecular genetic basis for the different viscosity behaviours in the genetically rough-seeded accessions is the r mutation and a consequence of the insertion in the starch-branching enzyme I-encoding gene (sbeI gene), affecting the carboxy-terminal region of the enzyme and the synthesis of amylopectin in developing pea seeds [35]. The mutation has been widely adopted by the vegetable industry and is the basis for most commercial vining cultivars.

Multivariate Analysis
Principal component analysis was performed including the peak and setback viscosity values; morphological and cooking parameters, namely, 100-seed weight, hydration capacity, unhydrated seeds, and cooking time; and composition in terms of percentage of protein, fibre, fat, and resistant starch (Figure 4).
The molecular genetic basis for the different viscosity behaviours in the genetically rough-seeded accessions is the r mutation and a consequence of the insertion in the starch-branching enzyme I-encoding gene (sbeI gene), affecting the carboxy-terminal region of the enzyme and the synthesis of amylopectin in developing pea seeds [35]. The mutation has been widely adopted by the vegetable industry and is the basis for most commercial vining cultivars.

Multivariate Analysis
Principal component analysis was performed including the peak and setback viscosity values; morphological and cooking parameters, namely, 100-seed weight, hydration capacity, unhydrated seeds, and cooking time; and composition in terms of percentage of protein, fibre, fat, and resistant starch (Figure 4). The first two components justified 51% (Component 1) and 13% (Component 2) of the total variance, accounting for 64% of the variance. The first component had a positive correlation with basic composition parameters, cooking time, and hydration capacity, while viscosity parameters showed a negative correlation with these parameters (as also observed in Table 2). The second component showed a negative correlation with 100-seed weight and a positive one with the unhydrated seeds parameter (Figure 4a)  The first two components justified 51% (Component 1) and 13% (Component 2) of the total variance, accounting for 64% of the variance. The first component had a positive correlation with basic composition parameters, cooking time, and hydration capacity, while viscosity parameters showed a negative correlation with these parameters (as also observed in Table 2). The second component showed a negative correlation with 100-seed weight and a positive one with the unhydrated seeds parameter (Figure 4a).
Accessions 220, 221, and 300 showed higher peak and setback viscosity values; accessions 263, 215, and 226 were separated from the group due to their higher percentage of unhydrated seeds; and accession 259 had higher percentages of fibre and fat (Figure 4b).
Moreover, it is possible to confirm that the 13 accessions phenotyped as rough but genetically classified as smooth behave similarly to all 64 accessions initially phenotyped as smooth, and that the 19 accessions confirmed as "rough" constitute a separate group along the first component.

Conclusions
The present results highlight the value of molecular analyses combined with the study of quality parameters, enabling the selection of appropriate pea germplasm and breeding for discrete end uses.
The pea collection analysed here displayed a favourable pasting profile for the development of flour for baking and other food formulations. Results for protein, fibre, and fat contents were comparable to those from other pulses. Resistant starch values varied greatly, however, among the pea accessions analysed; this component was negatively correlated to the pasting viscosity, an important contributor to cooking property variation.
Phenotype-based characterisation distinguished seeds according to shape, colour, and surface traits. While for the shape and colour classes defined, the results of physicochemical analyses were scattered, in contrast, all parameters differed significantly between the rough-and smooth-seeded classes. Of these, 29 rough-and 3 smooth-seeded accessions were further characterised for their allelic variation at the r and rb genetic loci. Indeed, 13 of the rough-seeded phenotyped accessions were genetically characterised as smooth, and their physicochemical responses were similar to the behaviour of the other smooth-seeded accessions.
A final PCA study was performed wherein the pea accessions were separated according to their surface type, linking this trait to cooking and to nutritional value traits, mainly determined by fibre, fat, and resistant starch composition.