Preliminary morphological characterization and evaluation of selected Bambara groundnut [Vigna subterranea (L.) Verdc.] genotypes for yield and yield related traits

Bambara groundnut [Vigna subterranea (L.) Verdc.; Syn: Voandzeia subterranea (L.) Thouars.] is an under-utilized grain legume grown in sub-Saharan Africa (SSA), mostly by women as a source of protein for subsistence (Ntundu et al., 2004). Bambara groundnut is third in importance in the SSA among grain legumes after groundnut (Arachis hypogea L.) and cowpea [Vigna unguiculata (L.) Walp.] (Sellschope, 1962; Rachie and Silvestre, 1977; Linnemann and Azam-Ali, 1993).
        
Bambara groundnut grains make up a complete balanced food (Rowland, 1993). The seeds are rich in protein (16-25%), carbohydrates (~ 63%) and oil (~18%). The predominant fatty acids in the oil include oleic acid, palmitic acid and linolenic acid (Minka and Bruneteau, 2000). Chemical analyses showed that the seed contains 32.50-32.72% of total essential amino acids including lysine, hisidine, arginine, leucine and isoleucine and 66.10-70.80% of the non-essential amino acids such as methionine, glycine, cysteine, tyrosine and proline (Minka and Bruneteau, 2000; Amarteifio et al., 2006; Aremu et al., 2006).
        
Various parts of Bambara groundnut are used for human consumption. Fresh seeds in pods may be boiled and eaten as a snack in a manner similar to boiled peanut, while dry seeds can be made into pudding (or steamed-paste) called Moi-Moi or Okpa (bean porridge) in some parts of Nigeria (Okpuzor et al., 2009), hence has the potential to enhance nutritional security for humans.
        
The crop has the potential of producing greater than 3,000 kg ha^-1 (Collinson et al., 2000). Major proportion of Bambara groundnut of the global production comes from West Africa, in which Nigeria is leading with a range of 45-50% (Goli et al., 1997).
        
Despite its various advantages, to date, the genetic potential of Bambara groundnut remains largely unexploited. Thus far, only farm level selection is being practiced whereby existing landraces consisting of several morhpo-types are evaluated and their seeds multiplied for use by farmers, hence the need to carry out within and between variations of such materials (Mohammed et al., 2016) to select desirable genotypes for further studies. As such there are no improved varieties of Bambara groundnut available in the major growing areas. Landraces can provide breeders with sources of genes for biotic and abiotic resistances, adaptability to different environments, nutritional characteristics and yield potential. In view of the above, this  study was aimed to characterize and evaluate for initial yield and yield component responses of 49 genotypes of Bambara groundnut derived from single plant selections made from a diverse germplasm collection using 26 morphological traits.

Plant material and study site
 
The 49 genotypes of the Bambara groundnut landraces used in this study are presented in Table 1. The study site was the Research and Training Farm of the University of KwaZulu-Natal, Pietermaritzburg, at Ukulinga, South Africa. The research was conducted in 2014 main season. The site is situated on a Latitude 30° 24’S, Longitude 29o 24’E and is 800 m above sea level.
 

 
Experimental design, field management and data collection
 
For both characterization and yield evaluation, a partially balanced lattice design was used with three replications and was randomised into the seven incomplete blocks across the three replications. The experimental plot comprised of three rows measuring 2.2 m x 3.0 m, with inter and intra row spacing of 0.4 m x 1.0 m. Sowing was done into a flat seedbed, with one seed sown to each stand. Missing stands were replaced within two weeks after sowing. All relevant agronomic practices were carried out to maintain a healthy crop.
 
Data on the 26 morphological (qualitative and quantitative) traits were generated on five plants selected from the central row of each plot within the incomplete blocks over the three replicates (IPGRI/IITA/BAMNET, 2000). The quantitative field data included number of days to 50% seedling emergence (SDE) by counting number of days from planting to 50% seedling emergence, while plant height (PHT), canopy spread (CNS), terminal leaf length (TLL), terminal leaf width (TLW) and petiole length (PETL) were measured using a measuring ruler and expressed in cm. Qualitative data recorded included leaf colour at emergence (LCE), terminal leave shape (TLS), growth habit (GH), stem pigmentation (SPG), petiole colour (PCL), leaflet joint pigmentation (this is the pigmentation at the point of attachment of petiolule to the petiole) (LJP), calyx colour (CCL), fresh pod colour (FPC), pod shape (PSP), dry pod colour (PCL), pod texture (PTX), seed shape (SSP) and seed eye pattern (SEY). The qualitative data were determined by visual observations.
        
Post-harvest quantitative data include dry biomass (BMA), pod weight (PDW), seed weight (SDW) were recorded in grams (g) using an OHAUS Precision Standard Measuring Scale, while hundred (100) seed weight (HSW) was recorded also in grams using a more sensitive Mettler Scale. Seed length (SDL), seed width (SDW), and seed height (SHT) were determined using a Digital Vernier Calipers (cm) on ten randomly, but well developed and uniform seeds taken from seeds used for 100 seed weight measurement, while qualitative data which were taken visually include kernel shape (PDS), kernel colour (PDC), kernel texture (PDT), seed shape (SDS) and seed eye pattern (SEY).
 
Data analysis
 
All the quantitative traits were subjected to analysis of variance based on the lattice procedure using Agrobase statistical software (Agrobase, 2005). Treatments’ means were separated by the least significant differences (LSD) at 5% probability. Cluster and principal component analyses were conducted on both quantitative and qualitative traits to determine similarities and dissimilarities among the genotypes using SPSS (SPSS, IBM Statistics 20). Using similarity matrix, a dendrogram was constructed to describe similarities and differences among the Bambara groundnut genotypes.

Morphological and yield attributes analyses
 
Genetic analysis diversity is one of the criteria for an organised sampling of breeding population/germplasm collection useful for the identification of desirable genotypes for hybridization (Razvi et al., 2018). The Bambara groundnut genotypes exhibited considerable variation among the morphological and seed traits. Highly significant (P<0.001) differences were detected for canopy spread, petiole length, weight of biomass, seed weight and seed height, while number of days to seedling emergence, pod weight, seed length and seed width were significantly (P<0.05) different (Table 2). In a similar diversity study using Bambara groundnut landraces in Tanzania, Ntundu et al., (2006) and Zenabou et al., (2014)  reported significant differences among quantitative traits including petiole length, plant spread, plant height, seed length and seed width, among others. In addition, variation in yield related traits have also been reported by Shegro et al., (2013), who showed that cultivar and environment may influence performance. These reports suggested that morphological and seed traits are useful for the characterization of Bambara groundnut and selection of desirable genotypes suitable for breeding, conservation and mass production. Razvi et al., (2018) showed that contribution due to character divergence vary from crop to crop.
 

 
Principal component analysis
 
Results of the principal component analyses (PCA) for the 26 agronomic and seed traits among the 49 Bambara groundnut genotypes are presented in Table 3. All 26 traits were grouped under nine components (Eigen values ≥ 1) which accounted for 79% of the variation. Leaf colour at emergence, petiole colour, leaflet joint pigmentation and calyx colour were highly correlated with PC₁, which accounted for 19.7% of the total variation. Seed traits which include seed length, seed width and seed height were correlated with PC₂, while pod weight and weight of biomass correlated with PC₃. Similarly, PC₄ contributed to 8.1% of the available variation and well correlated with terminal leaf length and plant height. Fresh kernel colour correlated positively with dry kernel colour in PC₅ contributing to 7.8% of the variation, suggesting that fresh kernel colour may affect kernel colour in dry condition. However, PCs 6, 7 and 8 had high correlations with 100 seed weight, kernel texture and leaf shape, contributing 6.2, 5.3 and 4.6% to the observed variability, respectively. PC₉ contributed to almost 4.0% of the variability in which stem pigmentation was important. In general, the PC analysis of the 26 traits indicated that PC₁ was composed of a number of traits that contributed for the greatest variation, followed by PC₂. In this study, it was observed that Bambara groundnut farmers may have driven the selection for specific morphological and seed traits. A Similar observation was made by Ntundu et al., (2006) who reported that leaf morphology, seed size and colour were morphological criteria used by farmers in Tanzania during selection.
 

 
Principal component biplot
 
The wide variation observed among the Bambara groundnut genotypes used in this study were further expressed by the PCA biplot (Fig 1). The biplot explained relationships and similarities that exist among the genotypes relative to the 26 measured traits. The genotypes scattered within the four quadrants produced by the PC₁ and PC₂ biplot. The genotypes showed a pairing orientation, irrespective of geographical locations within the axes, suggesting that they shared in common most of the features for the 26 traits that were studied. This feature of orientation would mean that movement of Bambara groundnut landraces across the African sub-region was indiscriminate. Using alfalfa germplasm (Medicago sativa L. subsp. sativa) (Yin et al., 2018) showed that genotype diversity was due to indiscriminate introduction of from various sources. Furthermore, pairing of the genotypes was associated more to individuals with possible common origin. Grouping of Bambara groundnut genotypes from the same region in Tanzania was earlier reported by Ntundu et al., (2006). Conversely, genotypes that scattered far apart within the axes would mean that they are distantly related with other genotypes within the same quadrant. PC₁ and PC₂ displayed 20% and 14% of the total variations in the quadrats. The results showed that the Bambara groundnut landraces have sufficient genetic diversity for selection and breeding purposes. Comparing the PC analysis and PCA biplot, the observed associations showed how the landraces share in common certain traits. Similar observations were made by Shegro et al., (2013) who suggested additional use of molecular markers to confirm such associations.
 

 
Cluster analysis
 
The degree of relatedness and differences among 49 genotypes for the 26 traits are presented in Fig 2. Four major cluster groups were identified, whereby Cluster I consist of three genotypes including two from Zambia (712-4 and 45-2) bearing Tan seed coat colour and one genotype (TV-93) from Sudan (acquired from IITA), which had a cream seed coat colour. The two Zambian genotypes were probably the same genotype, while the inclusion of the genotype from Sudan suggests that the three genotypes have a common origin or exhibit similarities in certain morphological features. The second cluster (Cluster II) was the largest, comprising of 24 genotypes distributed within two sub-clusters II a and II b. Cluster II a, consisted of 19 genotypes, while II b had five genotypes. The first sub-cluster IIa1 had an isolated genotype (TV-79-1) from IITA which originated from Kenya. Cluster II a6 had nine genotypes, with two forming a sub-sub-cluster (II a6-1) comprising of two genotypes (TV-14 and N212-4) from Ghana and Nigeria. Therefore, TV-14 and N212-4 may have come from the same ancestral origin. The other sub-sub-cluster, II a6-2 included six genotypes out of which four (211-68, 211-51, 211-88 and 211-90) were from CAPS, while the remaining two, M09-3-1 and M12-1 both originated from Zimbabwe. Since CAPS manages and sells Bambara groundnut seeds comprising of mixtures of landraces, the inclusion of the last two genotypes from Zimbabwe suggest that the accessions in this cluster may have had the same origin. Similar results were observed by Vyas et al., (2018) in cluster analysis using black gram [Vigna mungo (L.) Hepper] genotypes.
 

        
Cluster II b and II b1 consisted of only one genotype each, TV-27 and N212-5, respectively. TV-27 was from IITA and originates in Nigeria, while N212-5 originated from Kano in Nigeria, as well, suggesting that the two had unique origins in Nigeria. Cluster II b2-1 was made up of two genotypes (101-2 and 101-2-1) from Zambia. The genotype 101-2-1 was a selection from 101-2 and the two had in common their seed coat colour (cream stripe) (Table 1).
        
The divergence, within Principal component biplot and cluster analysis revealed that some genotypes share common origin, or certain traits among the 26 traits measured. In contrary, morphological diversity of landraces in Tanzania by Ntundu et al., (2006) showed that the genotypes were grouped according their regional collection zones, while  cluster grouping based on collection location was reported on cowpea in Ghana (Cobbinah et al., 2011).

The heterogenic nature with which the landrace collections exhibit would allow for two or more genoytpes to have been the same seed material, but bearing different names depending on where it was grown and/or sourced. Hence a concerted effort for further and advanced morphological and genomic characterization across Africa is important (Amadou et al., 2001). From the results in this study, best performing genotypes can be tested for mass production, breeding or germplasm conservation.