Relationship between seed traits and pasting and cooking behaviour in a pulse germplasm collection

Additional keywords: phenology, pulses.

Pulses are grain-legume crops harvested for dry grain and include beans, lentils (Lens culinaris), chickpeas (Cicer arietinum) and peas (FAO 1994). Their food use is increasingly supported by the promotion of healthy lifestyle and diets in order to prevent cancer, cardiovascular diseases, diabetes and obesity (Santos et al. 2017). The year 2016 was declared by the United Nations as the International Year of Pulses, with the aims of implementing a plan of action to increase awareness of the importance of legumes to human health, and increasing production and consumption of pulses (FAO 2015).

Latest FAO reports show that the production of pulses is increasing, with the latest data showing an annual production of 77 Mt worldwide (FAO 2014). Of this, dry beans represent ~33%, chickpeas 17%, dry peas 15% and lentils 7% (FAO 2016). Despite the trend of increasing worldwide production and considering pulses as protein crops, there is a gap between demand and supply, mainly in Europe (Häusling 2011).

With their gluten-free profile, high protein and fibre content, and low glycemic index, these grain legumes have lately been regarded as valuable alternative ingredients for the production of flours to be used in the development of new food formulations (Vaz Patto et al. 2015; Carbas et al. 2018). Examples include noodles (Song and Yoo 2017), pastas (Howard et al. 2011), biscuits (Sparvoli et al. 2016; Dauda et al. 2018) and cakes (Belghith-Fendri et al. 2016; Shaabani et al. 2018). As such, for quality improvement in breeding programs and for the selection of the best genetic material, certain traits must be characterised for genetic diversity within different germplasm collections. In chickpea collections, physical parameters such as size, shape and colour have been shown to affect variation in quality parameters including protein, fibre and fat content (Serrano et al. 2017). In addition, seed traits such as seed weight, hydration capacity and cooking time vary with chickpea genotype (Tripathi et al. 2012). However, significant effects of growing environment and storage conditions on physical and biochemical characteristics of grain legumes have also been reported (Sharma et al. 2014; Goyal et al. 2015).

Another advantage of legume seeds over cereal grains is their lower digestibility, associated with higher content of resistant starch (and lower glycemic index, referred to above) (Hoover and Zhou 2003). Starch pasting properties characterise the dissolution of starch and depend on physical and chemical characteristics of starch granules, being directly related to the development of viscosity (Atwell et al. 1988), and varying with the source of pulse starch and genotype under analysis (Byars and Singh 2016). These parameters affect the cooking and eating quality of the final products, so are key in the evaluation of legume flours as potential new ingredients in nutritionally improved formulations (Giuberti et al. 2015; Carbas et al. 2018).

The composition and structure of several pulse starches from specific varieties have been studied (Byars and Singh 2016; Carbas et al. 2018); however, little information is available on the diversity of rheological properties among accessions of large set of germplasm, which is necessary for the identification of outstanding accessions and their incorporation into breeding programs.

The goals of this study were to characterise the diversity of pasting and cooking properties among germplasm collections of faba bean (Vicia faba), chickpea, lentil and grass pea (Lathyrus sativus), and to identify outstanding accessions and understand the variation obtained by specific seed traits (e.g. colour, size, shape, surface and variety). Understanding the relationships between these factors, and how they influence each other, will contribute to the development of legume-based food products with desirable health benefits and sensory qualities.

For the study, 319 accessions were evaluated: 88 faba bean (accessions LEGVF 801–888), 86 chickpea (LEGCA 601–731), 47 lentil (LEGLC 401–509) and 98 grass pea (LEGLS 1–113). Faba bean accessions were selected from the germplasm breeding collection at IAS-CSIC (Córdoba, Spain), the CRF germplasm bank (Madrid), and the germplasm collection at INIAV-Oeiras, Portugal (WIEWS code PRT005). Chickpea accessions were selected from the germplasm breeding collection at IFAPA (Córdoba) and from the ICARDA germplasm bank. Lentil accessions were selected from the CRF germplasm bank. Grass pea accessions were selected from USDA (USA) and CRF germplasm banks and from the IAS-CSIC breeding collection. The accessions were multiplied in Córdoba during 2014, under the same field conditions, and irrigated and hand-weeded as needed. Harvest was performed by hand and seeds were stored at 5°C before analysis.

Classification of seed traits was based on information collected from different genetic resource databases, and classes were adapted according to the variability found for faba bean (Genesys Plant Genetic Resources 2017), chickpea (Integrated Breeding Platform 2014), lentil (Cristóbal et al. 2014) and grass pea (Rybiñski et al. 2008).

Faba bean seeds were classified according to seed dimensions (very small <10 mm by <13 mm, small 10–13 mm by 13–19 mm, medium 13–16 mm by 19–25 mm, large 16–20 mm by 25–30 mm, very large >30 mm by >20 mm) and colour (brown, dark brown, dark green, light brown, light green, mixed). Chickpea seeds were classified according to variety type (desi, kabuli), shape (angular, pea shaped, owl’s-head shaped), surface (smooth, rough, intermediate), seed size (very small <6 mm, small 6–8 mm, medium 7–9 mm, large 9–10 mm, very large >9 mm), and colour (black, brown, light brown, reddish brown, beige, light yellow, yellow brown, green, black brown, mosaic). Lentil seeds were classified according to size (small ≤3 mm, large >3 mm) and colour (brown, green, dark brown, mixed). Grass pea seeds were classified according to size (small <7 mm, medium 7–9 mm, large >9 mm), shape (triangular, rhomboid, round) and colour (cream, cream beige, cream red, cream white, cream green, brown, red brown, green brown, dark brown, brown grey, grey, mixed).

All four germplasm collections were evaluated for the following characteristics:

1. 100-Seed weight (g): 100 seeds from each harvested accession were randomly selected and weighted.

2. Hydration capacity (%): assessed by weighing 100 whole seeds and soaking for 16 h at room temperature. At the end of the soaking period, seeds were drained and reweighed to calculate weight difference, inferring the hydration capacity.

3. Unhydrated seeds: the seeds that remained entirely or partially unswollen after the 16-h soaking period.

4. Cooking time (min): estimated by the Mattson Cooker method (Ribeiro et al. 2007), using an adapted Mattson device with 32 pins of 98 g. Samples of 40 seeds were soaked in distilled water (volume twice the sample) for 16 h at room temperature before cooking. The soaked samples were then positioned in each of the 32 cylindrical holes in the Mattson device, so that the pins were in contact with the surface of the seeds. The Mattson device was placed into an electric rice cooker containing 2 L boiling water. Seeds were considered cooked when the tip of the brass rod passed through the seed. Cooking time was recorded in the passage of the 13th pin and stopped at maximum of 120 min.

Pasting properties were determined by using a rapid visco analyser according to an adaptation of the AACC 76-21 method (AACC 1999). Pasting analyses were conducted on duplicate flour samples (3 g in 25 mL water), held at 50°C for 1 min, heated at 12°C min^–1 to 95°C, held at 95°C for 2.5 min, cooled subsequently at 12°C min^–1 to 50°C, and held at 50°C for 3 min. The peak, trough, breakdown, final viscosity and setback from trough were expressed in centipoise (cP).

Contents of protein, fibre and fat were assessed with a near-infrared (NIR) analyser (MPA; Bruker, Billerica, MA, USA) with ground-flour calibrations for grain legumes provided by Bruker (n >500; R² > 90). NIR data were validated with 10% of the samples selected to cover the range and characterised by the following reference methods: protein by the combustion method ISO 16634-2:2016 (ISO 2016), calculated by multiplying nitrogen concentration by a conversion factor of 6.25; fat extracted by using petroleum ether in a Soxhlet apparatus according to ISO 6492 : 1999 (ISO 1999); and fibre by the intermediate filtration method ISO 6865 : 2000 (ISO 2000).

Data were analysed by using GraphPad Prism version 7.00 for Mac OS X (GraphPad Software, La Jolla, CA, USA; https://www.graphpad.com/). The variation within each seed trait was analysed by one-way analysis of variance, and Pearson correlation coefficients between traits were determined. Statistical significance was considered if P < 0.05.

Overall variation of physico-chemical, cooking, rheological and basic compositional characteristics within each species was assessed by principal component analysis (PCA), using Tanagra data mining software, version 1.4.5 (Lumière University Lyon 2, Lyon, France) (Rakotomalala 2005).

Average values of 100-seed weight, hydration capacity, unhydrated seeds and cooking time were calculated (Table 1). In general, faba bean seeds had the highest value of 100-seed weight (range 21–196 g) and lentils and grass peas the lowest (respective ranges 1.5–6 g and 6–37 g). Indeed, seed weight has been shown to be a morphological parameter that varies greatly among faba bean germplasms (Yahia et al. 2012; Backouchi et al. 2015). The mean hydration capacity for all four species varied in the range 100–120%, and the percentage of unhydrated seeds was highest in faba bean germplasm, reaching ~21%. Chickpea showed the highest values for cooking time, ranging from 16 to 120 min, and showing a large variability among genotypes, as previously reported (Tripathi et al. 2012). The small seed size of lentil prevented determination of its cooking time with the Mattson cooker.

Table 1.  Average ± residual standard deviation of seed parameters for the germplasm collections of faba beans, chickpeas, lentils and grass peas
n.d., Not determined

Pasting properties give important information about flour quality and consistency (Offia-Olua 2014), these being important predictors for new product sensorial analysis. Here, pasting properties varied between species; faba bean accessions had higher mean values for peak and trough viscosity (>2500 cP), breakdown (>250 cP), final viscosity (>4000 cP) and setback (>1500 cP) (Table 1). Lentil and grass pea accessions, as seen for 100-seed weight, showed the opposite behaviour to faba bean, with lower mean peak and trough viscosity (<700 cP) and final viscosity (<1100 cP). Because peak viscosity values are indicative of the strength of pastes and the extent of granule swelling (Liang and King 2003), these results suggest that lentil, grass pea and chickpea flours seem suitable to endure high-temperature processes in formulation of food products.

Correlation coefficients among physico-chemical and pasting properties were calculated for each species independently, in order to understand the relationships between properties within germplasm collections (Tables 2–5). In general, all pasting parameters were correlated, which was expected.

Table 2.  Correlation coefficients of different traits in the faba bean germplasm collection
FV, Final viscosity; SB, setback; SW, 100-seed weight; HC, hydration capacity; US, unhydrated seeds; CT, cooking time. *P < 0.05; **P < 0.01

Table 3.  Correlation coefficients of different traits in the chickpea germplasm collection
FV, Final viscosity; SB, setback; SW, 100-seed weight; HC, hydration capacity; US, unhydrated seeds; CT, cooking time. *P < 0.05; **P < 0.01

Table 4.  Correlation coefficients of different traits in the lentil germplasm collection
FV, Final viscosity; SB, setback; SW, 100-seed weight; HC, hydration capacity. * P < 0.05; **P < 0.01

Table 5.  Correlation coefficients of different traits in the grass pea germplasm collection
FV, Final viscosity; SB, setback; SW, 100-seed weight; HC, hydration capacity; US, unhydrated seeds; CT, cooking time. *P < 0.05; **P < 0.01

For the collection of faba bean germplasm (Table 2), negative Pearson correlations (P < 0.01) between final viscosity and seed weight, hydration capacity and cooking time were found. Higher seed weight correlated positively with higher protein content, hydration capacity and cooking time. Despite the tight association among seed size and cooking times, accessions LEGVF816, LEGVF822 and LEGVF823 exhibited high 100-seed weight (>120 g) and low cooking times (<50 min); therefore, those genotypes can be useful for breeding programs.

In the chickpea germplasm collection, significant negative correlations were found between protein, fibre and pasting properties, as well as a positive correlation of fibre and pasting properties with seed weight (Table 3).

As seen for faba bean seeds, a negative correlation was found (P < 0.01) between final viscosity values and hydration capacity in lentil seeds (Table 4) and grass pea seeds (Table 5). In grass pea seeds, no significant correlation was identified with cooking time.

In other studies, different physico-chemical properties such as composition, crystallinity and gelatinisation behaviour were shown to be linked to pasting properties of, for example, beans, lentils, chickpeas and pea seeds (Chung et al. 2008; Joshi et al. 2013; Carbas et al. 2018). In the present study, new traits were correlated with pasting properties, which assists in understanding the final rheological behaviour in flours of different legume species.

The means of peak, trough, break, final viscosity, setback, seed weight, hydration capacity, unhydrated seeds and cooking time for each seed trait according to germplasm collection are provided in appendices (Appendix 1–4). Given the correlation analysis presented above, final viscosity, seed weight, hydration capacity and cooking time were selected as the parameters that most vary, and are shown as the dependent variable in relation to each seed trait (Figs 1–4).

Fig. 1.  Mean ± s.e. values of final viscosity, hydration capacity, seed weight and cooking time for faba bean germplasm collection for each class of seed (a) size and (b) colour.

Fig. 2.  Mean ± s.e. values of final viscosity, hydration capacity, seed weight and cooking time for chickpea germplasm collection for each class of seed (a) variety, (b) shape, (c) size, (d) surface and (e) colour.

Fig. 3.  Mean ± s.e. values of final viscosity, hydration capacity, seed weight and cooking time for lentil germplasm collection for each class of seed (a) size and (b) colour.

Fig. 4.  Mean ± s.e. values of final viscosity, hydration capacity, seed weight and cooking time for grass pea germplasm collection for each class of seed (a) shape, (b) size and (c) colour.

In faba bean accessions, seeds were mainly distributed in small (n = 31), medium (n = 29) and large (n = 15) seed-size classes. As size increased, final viscosity values significantly decreased, whereas seed weight, cooking time and hydration capacity significantly increased (Fig. 1a). Considering the colour trait (Fig. 1b), there was a large amount of genetic variance, 32 of the 86 accessions categorised light brown, 14 light green, 13 mixed, 11 brown, 10 dark green, and six dark brown in colour. Seeds with light green colour seemed to show different behaviour with regard to seed weight and cooking time (P < 0.05) (Fig. 1b).

Chickpea germplasm was mainly divided into the two well-described varieties, desi (n = 26) and kabuli (n = 50). Kabuli seeds showed significantly higher seed weight (33 g v. 19 g for desi seeds), which is similar to previous findings, where kabuli seeds weighed 17.6 g and desi seeds 35.7 g (Tripathi et al. 2012). Kabuli seeds are usually described as having thinner seed coat (Serrano et al. 2017), and here, they had significantly higher final viscosity values (Fig. 2a). Besides being influenced by variety, final viscosity also significantly varied with seed shape (Fig. 2b), being higher in seeds of owl’s-head shape, whereas cooking time was much lower in these accessions. Bigger seeds had higher seed weight, as expected (Fig. 2c), and surface type did not affect the parameters under study (Fig. 2d). Most chickpea accessions were divided among beige (n = 27) and light yellow (n = 25) colour categories, and these two groups differed significantly for seed weight (Fig. 2e).

The lentil germplasm collection was divided between small (n = 20) and large (n = 27) seeds, which differed only in seed weight (Fig. 3a). With separation by seed colour, green lentil seeds had lower seed weight (Fig. 3b).

Among grass pea accessions, rhomboid seeds (n = 30) had higher final viscosity and seed weight values (P < 0.05), whereas round seeds (n = 9) had higher cooking times (P < 0.05) (Fig. 4a). Most grass pea seeds were of medium size (n = 43), and these significantly differed from large seeds (n = 23) in final viscosity and seed weight (Fig. 4b). As seen in chickpea seeds, grass pea accessions varied greatly in colour, the most prevalent colour being cream beige, which, alongside mixed colour (n = 10), had the lowest cooking time (Fig. 4c).

We included the compositional data relative to protein, fibre and fat contents in the PCA because there were several significant correlations. These three variables were added to peak, setback, 100-seed weight, cooking time, hydration capacity and unhydrated seeds, giving nine variables in total under analysis (Fig. 5).

Fig. 5.  Principal component analysis of score plot (left, with accession numbers) and loadings factors (right) for (a) faba bean, (b) chickpea, (c) lentil and (d) grass pea germplasm collections. Loading factors consisted of nine variables: peak, set back, cooking time, 100-seed weight, hydration capacity, unhydrated seeds, and contents of protein, fat and fibre. Note: the parameters cooking time and unhydrated seeds were not determined for lentil samples.

Our results demonstrate the inherent heterogeneity of faba bean germplasm, with wide phenotypic variation (Fig. 5a), as shown by others (Karaköy et al. 2014). The first and second principal components explained 36% and 18% of total variance, respectively, accounting for 54% of variance. On the factor-loading plot, we found that protein content was correlated with cooking time, hydration and 100-seed weight, and the positions of the accessions LEGVF841, LEGVF887, LEGVF873, LEGVF888, LEGVF886 and LEGVF884 allow inference about their high protein content and associated factors.

For chickpea accessions, variability by basic composition has been demonstrated (Serrano et al. 2017). Here, the resulting components of the PCA for chickpea explained 52% of total variance: 37% for the first component and 15% for the second (Fig. 5b). In general, fat content seemed more related to the viscosity parameters (peak and setback). Accession LEGCA604 presented the greatest hydration capacity, whereas accessions LEGCA608 and LEGCA609 had high values of protein and fibre content.

For the lentil germplasm collection, 54% of total variance was explained, corresponding to a first component of 31% and a second component of 23% (Fig. 5c). The viscosity parameters (peak and setback) appeared negatively correlated with protein content as well as hydration capacity (as already seen in Table 4). Hence, accession LEGLC423 could be isolated for its high fibre content and seed weight.

The grass pea PCA explained 44% of total variance, i.e. 24% (first component) and 20% (second component). Similar to observations for the lentil germplasm collection, and based on the first component, the viscosity parameters appeared negatively correlated with protein content and positively correlated with fat content and seed weight (Fig. 5d). Accession LEGLS022 separated from the rest of the group owing to its high 100-seed weight.

The results obtained in this study showed a high degree of variability between legume species for their pasting and cooking characteristics. With the exception of chickpea, seed weight was positively correlated with hydration capacity. Interestingly, in faba bean, no significant relation between seed traits and pasting properties and a negative correlation between final viscosity and cooking time was found, which is most important considering that the faba bean germplasm collection had the highest viscosity profiles.

In general, seed traits such as size, shape, colour, variety and surface type affected final viscosity, seed weight, hydration capacity and cooking time. Studying and understanding how these physico-chemical parameters vary with phenotypic and genotypic traits is important because they could be used as predicting factors for evaluation of germplasm cooking quality.

Longer cooking time can also be correlated with the chemical composition of seeds. It was possible to infer that cooking time was correlated with protein content in faba bean (which was also the legume with the highest seed weight), but in the other germplasm collections, protein and fibre content were more related to hydration capacity and percentage of unhydrated seeds.

In conclusion, taking into account the seed and flour characteristics mentioned above, the appropriate pulse can be selected, with higher viscosity profiles or lower cooking times, to substitute for cereal flours in development of, for example, pasta or noodles.

The authors declare no conflicts of interest.