Comparative Analysis of Agronomic Traits, Seed Color Variability and Protein Content in Pisum sativum L. Cultivars Across Diverse European Environments

Pea (Pisum sativum L.) is a legume rich in digestible protein and holds a prominent position as a crucial nutritional source across the globe. Depending on genotype and growth conditions, pea seeds can contain up to 33% protein (Dahl et al., 2012; Lam et al., 2018; Daba and Morris, 2021). Also, pea contains low levels of anti-nutritional compounds, providing a rich supply of amino acids and fiber (Clemente et al., 2015), with a unique ability for biological nitrogen fixation and mobilization of insoluble soil nutrients, bringing about qualitative changes in soil properties (Choudhary et al., 2023).
       
In the last fifty years, Europe has seen a significant rise in protein demand due to population growth. This has led to a decrease in animal husbandry, prompting the adoption of legumes as substitutes for animal proteins both in Europe and globally (Maurya et al., 2015). While pea cultivation has increased recently (Faostat, 2021), the demand for new protein sources continues to grow (Ismail et al., 2020). This increase is driven by high protein consumption in Europe, particularly plant-based protein, which exceeds recommended levels by an average of 70% (Westhoek et al., 2011). Peas, unlike soybeans, can be grown across a wider geographical range, including colder climates. Importantly, pea grain can be used directly in animal feed without prior heat treatment, which streamlines usage and reduces processing costs (Anderson et al., 2007).
       
Pea breeding aims to boost seed yield and protein content for wider applicability in diverse agricultural systems. To ensure stable and adaptable yields in peas, it’s important to evaluate how different pea varieties interact with their environment (G×E) across various agricultural conditions (Nassir and Ariyo, 2011; Haile 2020; Isik et al., 2017). Pea genotypes typically have two seed forms: smooth and wrinkled. While protein content and composition may differ between these forms, specific information remains inconclusive. Pea seeds vary in color, including brown, light and dark green, cream, or yellow, with pigmented, non-pigmented, or mixed varieties, depending on the genotype (Stjepanović et al., 2012). The protein content and composition may not be directly influenced by seed coat color, but there could be some differences related to morphological and physiological properties, such as seed surface or color (Guindon et al., 2021).
       
To address the growing demand for plant-based protein sources and recognize the paramount role of peas, this study investigates the protein content of pea cultivars with varying seed colors. The goal is to determine how different agroecological conditions impact pea seed protein content and agronomic traits related to seed yield.

Field trials were established in 2019 and 2020 at two distinct locations, the experimental field of the Institute of Field and Vegetable Crops, Rimski šančevi, Republic of Serbia (45°20'N, 19°51'E) and the experimental field of Agro Seed Research, Kessenich, Belgium (51°08'N, 5°48'E). The analyzed panel included research lines and varieties of dry and forage pea from Southern Europe. All material was kindly provided from germplasm collection from the Institute of Field and Vegetable Crops. The experiments were set up on 64 pea (Pisum sativum L.) genotypes, using the augmented design methodology within a row-column system (Cullis et al., 2006; Kumar et al., 2024) with four experimental blocks. Trials were implemented using the same experimental plan, with the difference of vegetative period from March until July for trial in Serbia and from April until August for trial in Belgium. The data for analyzed traits was measured and collected from ten randomly selected plants from each plot, avoiding marginal rows, or from a seed sample from the whole plot.
 
Weather conditions
 
Meteorological data from the Serbian trial plot (RHMZ, Annual Bulletin for Serbia in 2019 and 2020) revealed climate variations for both years. In 2019, the pre-sowing period had higher temperatures and less precipitation, balanced by above-average rainfall in April. May saw decreased temperatures and increased precipitation during pea bloom, while June experienced above-average precipitation during pod development, followed by an unusually dry July. In 2020, both temperature and precipitation exceeded averages during the pre-sowing period and March. April and May had below-average precipitation, while June saw significantly higher levels. The average temperatures from January to December 2020 were 0.6°C cooler than in 2019, with 61.1 mm precipitation, 8.4 mm more than in 2019, both above the long-term average (Fig 1, a). Meteorological data from the Belgian trial plot (Data Access Viewer - Agroclimatology) showed that, in 2019, March’s pre-sowing period had above-average temperature and precipitation, with below-average precipitation in April. June and July witnessed below-average temperatures during flowering and pod filling, respectively. Precipitation during the pea vegetative season (April to August) was below average, except in March and June. From May to August 2020, temperatures were below average, except for April, with above-average temperatures in the pre-sowing period (January and February). Precipitation variations occurred, with below-average levels in April and May. The average temperatures from January to December 2020 were 0.5°C warmer than 2019, with 47 mm precipitation, 2.2 mm less than 2019, both below the multi-year average (Fig 1, b).
 

 
Seed yield components
 
Analyzed agronomic traits were determined for each plot. Time to flowering (FT) and plot seed yield (PSY) were determined for the whole plots; pods per plant (PP), pod length (PL), seed weight per plant (SWPP) and seeds per pod (SPP) were determined on a random sample of 10 plants from each plot; a thousand seed weight (TSW) was determined regarding the method of seed weight determination (ISTA, 2015) and seed protein content (PC) was determined by Fourier transform near-infrared spectroscopy (FT-NIRS) on an Antaris™ II FT-NIR Analyzer.
 
Protein composition
The protein composition was determined by using SDS-polyacrylamide gel electrophoresis by Laemmli (1970). The results obtained by electrophoretic analyses were then quantified by gel scanning and densitometric analysis using BioVision (Wilber, Germany) software. The identification of 18 individual subunits was conducted based on molecular weight and information from previous studies by Barać et al., (2010) and Barać et al., (2011). The representation of each genotype’s subunits in the sum amounts to 100%.
 
Seed color and seed type
 
The determination was done by observing seed color within one month after harvest and seed type was observed immediately after harvest. Categorization of the seeds was done based on their color, resulting in the following groups: yellow (non-pigmented), green (pigmented and non-pigmented), mixed green (with varying shades of yellow), mixed non-pigmented (consisting of different shades of non-pigmented), mixed pigmented (consisting of different shades of pigmented) and mixed (a combination of pigmented and non-pigmented). Furthermore, the seeds were sorted into different types, namely smooth, wrinkled, a mixture of dimpled and smooth and a mixture of wrinkled and smooth (UPOV Pisum sativum TP TG/7/10 Rev. Pea).
 
Statistical analysis
 
Results from a two-year study were used to model spatial variation using a first-order autoregressive structure, while values for measured traits were predicted based on the spatial variation, using Best Linear Unbiased Predictions (BLUP). A mixed model analysis, following Demidenko et al., (2005), was conducted in Progeno software (Maenhout, 2018). Significant differences between average values across the two years were determined using a T-test. Descriptive statistics for both trials were computed using XLStat. Principal component analysis (PCA) was performed on traits from both sites, using mean values across both years, to examine the population structure of 64 pea genotypes, considering seed color and type, using Minitab (Minitab 17 trial version). Pearson’s correlation coefficients between BLUPs for quantitative traits were analyzed in R program version 4.2.2 (R Package “corrplot”, Wei and Simko, 2017), following Benesty et al., (2009). For non-parametric statistics comparing multiple groups of pea seeds based on color and type, a Kruskal-Wallis test was performed, according to Corder et al., (2009), with ggplot2 package in R program version 4.2.2 used for visual exploration of differences and relationships between seed groups (Wickham, 2016).

Agroecological impact on pea traits
 
In order to evaluate the impact of agro-ecological conditions on pea traits, the significance of the differences through the T-test was evaluated between the average values of two years for FT (days), PP, PL (cm), SPP, SWPP (g), TSW (g), SY (kg/ha) and PC (%). Based on the descriptive statistics shown in Table 1, it can be seen that FT had mean values of 22.3 and 14.9 days, mean PP values were 11.5 and 8.1, PL had similar values for both trials (5.8 cm for Serbian and 5.1 cm for Belgian trial); SPP was 8.7 and 6.6 in average, while SWPP was 7.2 g and 6.2 g for Serbia and Belgium, respectively. These traits showed significant differences between two different agroecological environments. Traits TSW, with average values of 182.6 g and 185.3 g and SY with averages of 104.9 kg/ha and 106.4 kg/ha showed no statistical differences between the two trial sites. Mean values for PC at the Serbian site exhibited an average protein content of 27.3%, while at the Belgian site, it was 27.4%, showing no impact of the environmental conditions on this trait. The selection of two geographically distinct locations aimed to enhance the study’s scope by considering diverse agroecological conditions. Findings suggest that imbalances in rainfall during the flowering phase can significantly affect interactions between plants and pollinators, thereby regulating flowering time (Kuppler and Kotowska, 2021). Insufficient rainfall, leading to heat stress as observed in the Belgian trial, results in pod rejection and reduces the number of pods per plant, in line with findings by Atung (2018) and Mohapatra et al., (2020). Additionally, unfavorable agroecological conditions reduce flowering time, contributing to decreased seed mass (Tawaha and Turk, 2004). This study concludes that considerable variations in agroecological conditions between the two investigated localities, particularly in rainfall patterns, significantly impact these characteristics. Aside from the differences in precipitation quantity and distribution, these features may also be influenced by the timing of sowing, according to Bozoglu et al., (2007). The research conducted by Barcchiya et al., (2018) and Saxesena et al., (2014) suggests that SPP is mainly influenced by environmental factors rather than genotype, although it does exhibit significant heritability in a broader sense, while PL is a heritable trait, meaning that genetic factors, more than agroecological conditions, have a major influence on this trait (Avci and Ceyhan, 2006), which can be seen from similar results on two localities. By comparing the average of the two-year results at both locations, very small differences were observed in the TSW, which confirms that this trait is highly heritable and genetically determined (Burstin et al., 2015; Georgieva et al., 2016; Singh et al., 2017). Similar can be said for seed protein content, which aligns with the previous research conducted by Crosta et al., (2021). The yield is significantly influenced by the interaction of genetic and environmental factors, where cultivation at high temperatures caused by climate change, as well as the amount and distribution of precipitation, contribute to low yield (Acikgoz et al., 2009; Atung, 2018).


       
Pearson’s correlation coefficients were calculated to evaluate the relationship between traits, using average values from two years (Fig 2). A significant positive correlation was expressed between SY and TSW (0.99) and SPP expressed a positive correlation with PP (0.70) and PL (0.52). Also, a positive correlation was expressed between PL and two traits, TSW and SY (both 0.65). A significant negative correlation was expressed between PC and two traits, TSW (0.66) and SY (0.65). The correlations between the rest of the pairs were of weak or no significance, indicating possible non-linear interactions. A strong positive correlation between SY and TSW is to be expected, this trait is directly related to seed yield and they are mostly positively correlated (Khan et al., 2017). On the other hand, TSW is negatively correlated with PC content and the same was observed between SY and PC, which is negative in most cases (Dhama et al., 2010; Mohanty et al., 2020; Asha et al., 2020). TSW had a positive influence on SY, possibly because heavier seeds provide a more favorable condition for the growth of plants that are highly adaptable to agroecological growing conditions. PC, on the other hand, was negatively correlated with SY, indicating a trade-off between protein content and yield. The results of the correlation of PL, which was positively correlated with SPP, TSW and SPP, were similar to Naeem et al., (2020), while the positive correlation between SPP and PP were contrary to the findings of Mukherjee et al., (2023).
 

 
Understanding protein composition variability in pea genotypes
 
PCA analysis has been used to gain insight into the similarities and differences among pea varieties based on their physicochemical composition (Guindon et al., 2021), basic composition variability (Santos et al., 2019), genetic diversity (Ouafi et al., 2016) and bioactive compounds (Han et al., 2023). Multivariate analysis was performed based on the examination of mean values of all traits for both locations, in order to investigate the population structure of 64 pea genotypes differing in color and seed type. The PCA analysis of the tested pea genotypes for the first two main components is graphically represented (Fig 3). The results of the multivariate analysis show the separation of pigmented seeds from mixed non-pigmented seeds by the first axis (35.9%) and yellow non-pigmented seeds by the second axis (17.9%), with no clear grouping concerning seed type. The results obtained from multivariate analysis in this study indicate a clear distinction between yellow pea seeds and both pigmented and non-pigmented green seeds, with a noticeable separation of mixed pigmented genotypes. In contrast, there was no clear structure to the PCA plot in the analysis based on the division of genotypes by seed type. Gixhari et al., (2014) have supported these outcomes with their research on pea, employing PCA to demonstrate that traits such as seed number per pod, weight of 1000 seeds and genotype yield account for a significant portion of the observed variability. Also, this could be related to morphological and physiological properties, similar to the findings of Guindon et al., (2021), where yellow varieties showed superior values for weight and seed size, which affect the number of seeds per pod, while wrinkled varieties showed higher protein content.
 

       
An additional analysis was carried out to investigate variations in protein composition by categorizing pea genotypes based on color (Fig 4, a) and seed type (Fig 4, b). There were variations in the ratio of vicilin/legumin among seeds of various colors, although they were not statistically significant (p-value = 0.9963 according to the Kruskal-Wallis test). According to the visual representation of the results (Fig 4, a), the green non-pigmented seed exhibited the highest concentration of vicilin and the green pigmented seed exhibited the highest concentration of α legumin. The content of convicilin and α legumin was similar across all varieties of seeds, except for genotypes with mixed pigmented seed, which exhibited the lowest level of convicilin content. Concerning seeds of various types, there were no significant variations in protein content among them (p-value = 0.9778 according to the Kruskal-Wallis test), except for genotypes with mixed seed (smooth/wrinkled), which exhibited the highest concentration of convicilin and the lowest concentration of α legumin (Fig 4, b). The protein composition differed between seeds of different colors, showing more similarities between seeds of different types. Mainly, yellow seeds, which are mostly associated with dry pea genotypes, are considered to have different protein content, compared to green or pigmented seeds, which most often belong to forage or vegetable peas, as well as wild pea relatives. Genotypes with dark seed color exhibited the lowest mean vicilin content compared to other genotypic groups, as indicated by the percentage distribution of protein subunits. Conversely, genotypes with green pigmented seeds displayed the highest α legumin content compared to the observed groups. These findings could bear significant implications for both industrial processes and food production. This is especially relevant due to the emulsification properties of vicilin, which render pea genotypes with higher vicilin content valuable for technological processing (Barać et al., 2010). Moreover, genotypes with elevated legumin content offer considerable nutritional value, as confirmed by Villa et al., (2018). Notably, Gabriel et al., (2008) reported a negative correlation between amino acid digestibility and legumin levels, while Rubio et al., (2014) found that vicilin fractions are rich in arginine, isoleucine, leucine and lysine compared to the leguminous fraction. It is worth mentioning that genotypes with darker seed color (mixed pigmented) have the lowest representation of α and β legumin subunits.
 

In conclusion, our research emphasizes the profound influence of agroecological conditions on pea traits. We discovered that variations in rainfall and other environmental factors between the two examined localities had a significant impact on the duration of flowering, the number of pods and seeds and also seed weight per plant. The findings underscore the importance of understanding the interplay between environmental conditions and plants, which in turn offer insights and valuable guidance for agricultural practices and the optimization of crop yields in diverse agroecological settings.
       
Shifting the focus to protein composition, our research shed light on notable disparities among pea seeds of different colors and types. Particularly, genotypes with dark seed color exhibited the lowest vicilin content, while those with green pigmented seeds displayed the highest á legumin content. These variations carry significant implications for the food industry and nutrition. Vicilin content enhances technological processing, while legumin content enriches nutritional value. Our findings hold promise for informing the potential use and diverse applications of different varieties of pea genotypes.

This project has received funding from the European Union’s Horizon 2020 Programme for Research and Innovation under grant agreement no. 727312 (project EUCLEG) and the Ministry of Education, Science Technological Development and innovation of the Republic of Serbia, grant number: 451-03-66/2024-03/200032.

All authors declare that they have no conflicts of interest.