Introduction
Yam is a popular staple in West Africa (Asiedu and Sartie, Reference Asiedu and Sartie2010; Darkwa et al., Reference Darkwa, Olasanmi, Asiedu and Asfaw2020a). Its value chain sustains ~5 million people's livelihood, including the rural farming households, traders, transporters and processors (Mignouna et al., Reference Mignouna, Akinola, Abdoulaye, Alene, Manyong, Maroya, Aighewi, Kumar, Balogun, Lopez-Montes, Rees and Asiedu2020). Yam is an essential source of carbohydrates, vitamins, essential minerals, fibres, with a low glycemic index (Akinola et al., Reference Akinola, Ezeorah and Nwoko2019). Benin is part of the African yam belt, a region producing more than 90% of the world yam production. It produces ~2.9 million tons of yam annually and ranks fourth after Nigeria, Ghana and Côte d'Ivoire (FAO, 2020). Yam is a primary staple food and symbolic in many key socio-cultural and religious events (Zannou et al., Reference Zannou, Ahanchédé, Struik, Richards, Zoundjihékpon, Tossou and Vodouhè2004; Loko et al., Reference Loko, Fagla, Orobiyi, Glinma, Toffa, Koukoui, Djogbenou and Gbaguidi2019). The country produces several yam species, including Dioscorea cayenensis, D. rotundata, D. alata and D. dumetorum (Scarcelli et al., Reference Scarcelli, Tostain, Vigouroux, Agbangla, Daïnou and Pham2006). Among these species, D. rotundata is the most popular and preferred for its economic profitability and suitability for a range of local food recipes.
As part of the crop improvement and genetic conservation efforts, several genetic diversity studies have been conducted on yam across the globe in general and in the sub-region in particular. These include an inventory and characterization of local cultivars using phenotypic traits (Dansi et al., Reference Dansi, Zoundjihekpon, Mignouna and Quin1997; Loko et al., Reference Loko, Dansi, Linsoussi, Tamo, Vodouhè, Akoegninou and Sanni2013, Reference Loko, Adjatin, Dansi, Vodouhe and Sanni2015; Etchiha et al., Reference Etchiha, Affokpon, Loko, Agbangla and Dansi2019), random amplified polymorphism DNAs (RAPD) and simple sequence repeat (SSR) molecular markers (Dansi et al., Reference Dansi, Mignouna, Zoundjihékpon, Sangare, Ahoussou and Asiedu2000; Tostain et al., Reference Tostain, Agbangla, Scarcelli, Mariac, Daïnou, Berthaud and Pham2007; Missihoun et al., Reference Missihoun, Ahanhanzo, Agbangla, Dansi and Akpagana2009; Loko et al., Reference Loko, Ranjana, Agre, Dossou-Aminon, Orobiyi, Djedatin and Dansi2017). Although a high genetic diversity was reported in Benin, only a few cultivars are widely grown for their superior agronomic performance and market value (Dansi et al., Reference Dansi, Mignouna, Zoundjihékpon, Sangare, Ahoussou and Asiedu2000; Etchiha et al., Reference Etchiha, Affokpon, Loko, Agbangla and Dansi2019). Besides, yam yield in this country is still meagre (13 t/ha) compared to the crop's yield potential in the region (40–50 t/ha) (Frossard et al., Reference Frossard, Aighewi, Aké, Barjolle, Baumann, Bernet, Dao, Diby, Floquet, Hgaza and Ilboudo2017). Developing improved varieties combining high tuber yield and superior food quality traits is an effective strategy to enhance yam productivity and increase farmers’ interest in yam farming (Darkwa et al., Reference Darkwa, Olasanmi, Asiedu and Asfaw2020a). An effective yam improvement programme through breeding, to develop such varieties, requires proper knowledge of the genetic resources available to guide the selection of cross-compatible and suitable parents (with desirable traits) to be involved in hybridization (Mondo et al., Reference Mondo, Agre, Edemodu, Adebola, Asiedu, Akoroda and Asfaw2020). As above-mentioned, both molecular and phenotypic analyses have been routinely used in yam for cultivar characterization prior to hybridization to achieve breeding objectives. However, there are increasing evidences on the limitations of each of these characterization approaches.
Molecular markers previously used for yam genetic diversity studies include RAPD (Asemota et al., Reference Asemota, Ramser, Lopez-Peralta, Weising and Kahl1995), amplified fragment length polymorphism (AFLP) (Mignouna et al., Reference Mignouna, Ellis, Knox, Asiedu and Ng1998, Reference Mignouna, Mank, Ellis, Van Den Bosch, Asiedu, Ng and Peleman2002a, Reference Mignouna, Mank, Ellis, Van Den Bosch, Asiedu, Abang and Peleman2002b; Terauchi and Kahl, Reference Terauchi and Kahl1999), SSR (Arnau et al., Reference Arnau, Némorin, Maledon and Abraham2009; Girma et al., Reference Girma, Gedil and Spillane2017; Loko et al., Reference Loko, Ranjana, Agre, Dossou-Aminon, Orobiyi, Djedatin and Dansi2017; Mulualem et al., Reference Mulualem, Mekbib, Shimelis, Gebre and Amelework2018; Babil et al., Reference Babil, Yamanaka, Girma, Matsumoto, Tamiru-Oli, Bhattacharjee, Abberton, Muranaka, Asiedu, Terauchi and Takagi2021), inter SSRs (Ousmael et al., Reference Ousmael, Tesfaye and Hailesilassie2019) and single nucleotide polymorphism (SNP) markers (Agre et al., Reference Agre, Asibe, Darkwa, Edemodu, Bauchet, Asiedu and Asfaw2019; Bhattacharjee et al., Reference Bhattacharjee, Agre, Bauchet, De Koeyer, Lopez-Montes, Kumar, Abberton, Adebola, Asfaw and Asiedu2020; Darkwa et al., Reference Darkwa, Agre, Olasanmi, Kohtaro, Matsumoto, Powell, Bauchet, De Koeyer, Muranaka, Adebola, Asiedu, Terauchi and Asfaw2020b). However, the reproducibility and reliability of markers like AFLP are limited, time-consuming and have low distribution across the genome. From the last decade, SNPs generated through the Next Generation Sequencing approaches (Diversity Arrays Technology (DArT) and genotyping-by-sequencing) are the most widely used in yam diversity studies because of their stability and abundance in the genome (Girma et al., Reference Girma, Hyma, Asiedu, Mitchell, Gedil and Spillane2014; Akakpo et al., Reference Akakpo, Scarcelli, Chaïr, Dansi, Djedatin, Thuillet, Rhoné, François, Alix and Vigouroux2017; Siadjeu et al., Reference Siadjeu, Mayland-Quellhorst and Albach2018; Scarcelli et al., Reference Scarcelli, Cubry, Akakpo, Thuillet, Obidiegwu, Baco and Vigouroux2019; Bhattacharjee et al., Reference Bhattacharjee, Agre, Bauchet, De Koeyer, Lopez-Montes, Kumar, Abberton, Adebola, Asfaw and Asiedu2020; Darkwa et al., Reference Darkwa, Agre, Olasanmi, Kohtaro, Matsumoto, Powell, Bauchet, De Koeyer, Muranaka, Adebola, Asiedu, Terauchi and Asfaw2020b).
According to Andrade et al. (Reference Andrade, Barbosa and Pereira2017); Alves et al. (Reference Alves, Silva, De, Albuquerque and Santos2017); Agre et al. (Reference Agre, Asibe, Darkwa, Edemodu, Bauchet, Asiedu and Asfaw2019); Nkhoma et al. (Reference Nkhoma, Shimelis, Laing, Shayanowako and Mathew2020); Stanley et al. (Reference Stanley, Menkir, Agre, Ifie, Tongoona, Unachukwu, Meseka, Mengesha and Gedil2020), genotypic and phenotypic data display very low or negligible correlations and produce non-duplicate information. A large portion of variation detected by molecular markers is non-adaptive compared with phenotypic characters which are influenced by the environment. In yam, the effectiveness of combining phenotypic and molecular marker information for dissecting genetic diversity has been reported (Sartie et al., Reference Sartie, Asiedu and Franco2012; Agre et al., Reference Agre, Asibe, Darkwa, Edemodu, Bauchet, Asiedu and Asfaw2019; Darkwa et al., Reference Darkwa, Agre, Olasanmi, Kohtaro, Matsumoto, Powell, Bauchet, De Koeyer, Muranaka, Adebola, Asiedu, Terauchi and Asfaw2020b). Studies in yam and other crops suggest that combined analysis for phenotypic and genotypic information is useful for assessing the functional genetic diversity in crop plants and for minimizing limitations from either analysis approach. Genetic diversity assessment studies on Beninese white Guinea yam cultivars were conducted independently using either phenotypic or genotypic data. However, a joint analysis has not yet been deployed to explore the diversity of this crop in Benin. Hence, this study's objective was to assess the genetic structure and diversity in a panel of farmers’ cultivars across the country using both phenotypic and SNP markers. Our study therefore complements previous studies in the country that used either morphological or molecular data alone. Proper genetic resource characterization provided by this joint analysis will be instrumental for efficient parental selection for yam breeding programmes in West Africa, including Benin. Besides, it will strengthen the crop genetic resource conservation efforts which were previously relying on either phenotypic or genotypic information alone.
Materials and methods
Collection of plant materials, field establishment and phenotyping
A total of 88 white Guinea yam (D. rotundata) cultivars were collected through an ethnobotanical survey in 20 villages and 16 markets of Benin's four largest yam production regions (Atacora, Donga, Borgou, Collines) (online Supplementary Fig. S1). The collection targeted the most popular cultivars with high market value to be used for yam crop improvement. The farmers’ preference criteria for tuber yield potential, tolerance to weed competition, drought and diseases, long shelf life, good taste and aroma of processed food products and slow tuber flesh oxidation were considered in the sampling process.
These varieties were planted at the experimental farm of the Laboratory of Biotechnology, Genetic Resources and Plant and Animal Breeding (BIORAVE) located at Massi, Zogbodomey district, during the 2019 growing season (February–November). The field was established using an 11 × 8 lattice design with two replications. The plot consisted of an 11 m column with five plants per variety, at a spacing of 1 m × 1 m. No fertilizer was applied, while manual weeding was done regularly to maintain the plot free of weed. A total of 53 agro-morphological variables covering different plant parts (leaves, stem, flowers, fruits, roots, tubers) (online Supplementary Table S1) were assessed following the procedures described in the yam crop ontology (Asfaw, Reference Asfaw2016).
Genotyping
About 10 g of young, healthy and fully expanded leaves from each genotype were collected using silica-gel granules with a colour indicator. Leaves were stored in the silica-gel for 72 h to remove the moisture. Subsequently, DNA extraction was carried out at the Bioscience centre, International Institute of Tropical Agriculture (IITA), Ibadan, Nigeria, using the CTAB procedure with slight modification (Dellaporta et al., Reference Dellaporta, Wood and Hicks1983). The DNA quality was ascertained by running the gDNA in a 1% agarose gel while NanoDrop 2000 spectrophotometer was used to estimate its concentration and purity.
DArT genotyping was performed as described by Sansaloni et al. (Reference Sansaloni, Petroli, Carling, Hudson, Steane, Myburg and Kilian2010). For the sequencing-based DArT genotyping, complexity reduction methods optimized for yam at the DArT Pty Ltd., Australia, were used. PstI_ad/TaqI/HpaII_ad with TaqI restriction enzyme was used to eliminate a subset of PstI–HpaII. The pstI-site-specific adapter was tagged for the 88 accessions with different barcodes enabling encoding a plate of DNA samples to run within a single lane on an Illumina GAIIx.
Data analysis
Phenotypic analysis
Prior the Least Squares means (LSmeans) estimation, the data were scaled and the LSmeans generated for the 88 yam varieties were used for principal component analysis (PCA) with FactoMineR (Le et al., Reference Le, Josse and Husson2008) and missMDA (Josse and Husson, Reference Josse and Husson2016). The optimum number of discriminant factors was determined through Peres-Neto et al. (Reference Peres-Neto, Jackson and Somers2003) principle by considering factors with Eigenvalues >1. On each factor, the variables were declared discriminative and retained for subsequent analysis with a correlation above 0.5. Retained variables were then subjected to correlation plot using corrplot function implemented in R. Considering the discriminant variables, Gower dissimilarity matrix was generated using cluster package implemented in R. Count and nominal and or categorical variables (plant vigour, sprout colour, plant type, stem number per plant, number of internodes, spines on stem above base, flowering degree/intensity, plant sex, inflorescence type, tuber size, number of tubers harvested per plant, number of marketable tubers, number of tubers damaged by rot and disease, tuber shape, tuber skin thickness, tuber flesh oxidation, flesh colour, absence/presence of corms on tubers) were listed in the model and transformed using log ratio function prior generating the dissimilarity matrix. The final hierarchical cluster (HC) was then performed based on ward.D2 method in cluster R package (Maechler et al., Reference Maechler, Rousseeuw, Struyf, Hubert and Hornik2021). Gower dissimilarity matrix was subjected to various diversity indexes such as Shannon–Wiener Index (H′), Inverse Simpson's (HB), Simpson's Index (λ) and Pilou evenness (J) using vegan library (Oksanen et al., Reference Oksanen, Guillaume, Friendly, Roeland, Dan, Minchin, O'Hara, Simpson, Solymos, Henry, Szoecs and Wagner2019). The reason for using different indices for diversity assessment was solely for increasing the reliability of conclusions.
SNP analysis
For quality control, DArTseq SNP-derived markers were filtered to remove the unwanted SNP markers using the software PLINK 1.9 and VCFtools. Markers and genotypes with >20% missing data were eliminated. Rare SNPs with <5% minor allele frequencies and low coverage read depth (<5) were also removed. In the end, only informative 9725 DArT-SNP markers and 88 cultivars were used for the subsequent analysis.
The summary statistics such as observed and expected heterozygosity, minor allele frequency (MAF) and polymorphic information content (PIC) were estimated using VCFtools (Danecek et al., Reference Danecek, Auton, Abecasis, Albers, Banks and DePristo2011) and PLINK 1.9 (Purcell et al., Reference Purcell, Neale, Todd-Brown, Thomas, Ferreira, Bender and Sham2007). Mutation transversion and transition were determined using the SniPlay web base (Purcell et al., Reference Purcell, Neale, Todd-Brown, Thomas, Ferreira, Bender and Sham2007). Dosage SNP format (0, 1, 2) was generated using recodeA function implemented in Plink where 0 is the homozygote reference, 1 the heterozygote and 2 the homozygote alternative. Dosage format was then subjected to Jaccard dissimilarity matrix using phylentropy R package (Drost, 2018). Jaccard dissimilarity matrix was then used to estimate the genetic diversity indexes such as Shannon–Wiener Index (H′), Inverse Simpson's (HB), Simpson's Index (λ) and Pilou evenness (J) to quantify the level of genetic diversity countrywide in Benin as well as within and among surveyed geographical zones using the vegan library (Oksanen et al., Reference Oksanen, Guillaume, Friendly, Roeland, Dan, Minchin, O'Hara, Simpson, Solymos, Henry, Szoecs and Wagner2019).
Besides, three multivariate analyses, including admixture for population structure, PCA, and cluster analysis with Ward method, were employed. Binary file generated from vcf file was then subjected to admixture analysis using ‘adegenet’ R package (Jombart et al., Reference Jombart, Devillard and Balloux2010). The optimal number of clusters was inferred using k-means analysis after varying the number of clusters from 2 to 40. Through the admixture analysis, genotypes with membership proportions (Q-value) ≥60% were assigned to groups. In comparison, those with membership probabilities <60% were designated as admixtures (Salazar et al., Reference Salazar, González, Araya, Mejía and Carrasco2017). For the HC analysis, generated Jaccard dissimilarity matrix was used and the HC was plotted using Ward.D2 method. Further, PCA was conducted to reveal the genetic relationships among the yam varieties using GenAlEx v. 6.503 software (Peakall and Smouse, Reference Peakall and Smouse2012) by considering the collection zone as a factor. Jaccard dissimilarity matrix was further subjected to the analysis of molecular variance (AMOVA) using GenAlEx v. 6.503 (Peakall and Smouse, Reference Peakall and Smouse2012) to partition components of genetic variance among and within the populations (k).
Joint analysis of phenotypic and molecular data
The Gower and Jaccard dissimilarity matrices from both phenotypic and molecular data were combined to define potential genetic groups. The joint matrix was calculated as the algebraic sum of the Gower and Jaccard matrix generated through the phenotypic and the SNP marker data, respectively (Alves et al., Reference Alves, Bhering, Rosado, Laviola, Formighieri and Cruz2013; Andrade et al., Reference Andrade, Barbosa and Pereira2017). Mantel test was performed using the Monte-Carlo method with 9999 permutations to assess the correlation among the three matrices (phenotypic, genotypic and the combined). Phenotypic data of the different groups obtained through the joint analysis were subjected to k-mean analysis and the performance of each cluster was compared by Rstatix package (Kassambara, Reference Kassambara2020).
Results
Phenotypic diversity
The first 14 principal components were identified as most discriminative and accounted for 80.78% of the cultivars’ total phenotypic variation. Thirty (30) traits were correlated to the first 14 principal components (online Supplementary Table S2). The first principal component explained 13.98% of the total variation and was significantly associated with tuber size (TBRSZ), the number of marketable tubers (NMT), marketable tuber weight (MTW), marketable tuber length (MTLS), marketable tuber width (WMT) and tuber yield. The second principal component accounted for 10.16% of the total variation and was significantly correlated with six variables (online Supplementary Table S1). The third component explained 7.62% of the total variation and was positively associated with four phenotypic traits. Number of rotten tubers (NRDB) and weight of rotten tubers (RDTW) positively contributed to the fourth principal component, which explained 7.16% of the total variation.
Phenotypic correlation among the discriminating variables revealed high and positive correlations among total tuber yield, number of tubers, tuber weight, plant vigour, tuber length and tuber width while a negative correlation was recorded between the total tuber yield and tuber flesh oxidation (online Supplementary Fig. S2). Diversity indexes based on Gower dissimilarity matrix revealed the presence of moderate genetic diversity across the country (Table 1).
Table 1. Genetic diversity indices based on SNP data across collection zones of Benin
Inverse Simpson's (HB); Shannon–Wiener Index (H′); Pilou evenness (J); Simpson's Index (λ).
Cluster analysis based on phenotypic traits differentiated the 88 varieties into two major clusters (Fig. 1). The first cluster (red) was composed of 62 yam varieties with intermediate to late maturity cycle, moderate flowering intensity, moderate tuber yield (4.10 kg per plant) and high tuber flesh oxidation. The second cluster (green) comprised 26 early maturing yam varieties with low flowering intensity and good agronomic performance. Varieties in this cluster had high tuber yield (7.96 kg per plant) and high marketable tuber weight.
Fig. 1. Phenotypic characterization of Beninese white Guinea yam varieties based on Gower dissimilarity matrix using the 30 most discriminating traits. The colours represent the two clusters: Cluster 1 (red) and cluster 2 (green).
Molecular diversity and population structure
A total of 20,000 SNP markers was initially generated, from which 9725 were retained after the removal of low-quality SNP markers. The SNP markers were unequally distributed across the 20 yam chromosomes with a minimum of 212 SNPs on chromosome 13 to a maximum of 1469 SNPs on chromosome 5 (online Supplementary Fig. S3a). The SNP mutation showed that transition (T.S.) was greater (59%) than the transversion (T.V.) in the yam genome. Among the transition mutations, A/G had the highest occurrence rate in the genome (online Supplementary Fig. S3b). In contrast, for the transversion mutations, A/C appeared at the highest rate. Values of 0.22 and 0.24 were recorded as averages for the observed and expected heterozygosity, respectively. The PIC varied from 0 to 0.37, with an average of 0.20. The MAF was 0.17. Using the 9725 SNP markers, all cultivars evaluated were estimated to be diploid.
The Jaccard dissimilarity matrix displayed a high range of genetic distance from 0 to 0.89. Hierarchical clustering-based DArT-SNP marker grouped the 88 yam varieties into three major clusters (Fig. 2). Five cultivars were grouped in cluster 1 (red) (Fig. 2) with the lowest genetic distance (0.10) obtained between Taatimanin and Gakatele. In this cluster, the highest genetic distance (0.21) was observed between Laboko and Taatimanin. Cluster 2 (green) had the highest membership (68), with its members widely distributed across the different yam agro-ecologies. In this cluster, member cultivars were majorly identified as males except Gnidou, Katala, Eguede and Sotoboua which had female flowers. Cluster 3 (blue) was made of 15 cultivars (Fig. 2) and displayed 0.22 as an average genetic distance within the cluster. Members of this cluster were identified as females with low to high flowering intensity.
Fig. 2. Assessment of the genetic diversity of the 88 Beninese white Guinea yam varieties based on Jaccard dissimilarity matrix using the DArT-SNP markers. The colours represent the three clusters: Cluster 1 (red), cluster 2 (green) and cluster 3 (blue).
Through the Bayesian Information Criteria (BIC) analysis, population structure revealed the presence of deflection at k equal to 3 (online Supplementary Fig. S4, Fig. 3). Using a 60% membership probability threshold, 74 cultivars (84.09%) were successfully assigned to the three clusters. In contrast, 14 cultivars with an association probability of <60% were designated as admixtures (Fig. 3).
Fig. 3. Graphical representation of the 88 Beninese yam varieties’ population structure based on admixture analysis. Subpopulations were set at k = 3. The colours represent the three clusters: Cluster 1 (red), cluster 2 (green) and cluster 3 (blue) based on a membership coefficient of ≥60%.
Shannon–Wiener, Inverse Shannon, Simpson and Pilou diversity indices revealed the presence of moderate yam genetic diversity in Benin (Table 1). The level of genetic diversity was relatively high in Collines, followed by Borgou and Atacora regions. At the same time, the lowest index was recorded in Donga (Table 2).
Table 2. Pairwise fixation index (F st) values among the genetic groups
The dispersion of the genotypes on the first two principal components, which explained 44.2% of the total molecular variation, showed that the 88 yam varieties were clustered irrespective of their geographical origins (online Supplementary Fig. S5). The pairwise fixation index (F st) value further confirmed the low genetic differentiation between geographical origins (Table 2). The AMOVA also revealed high genetic variability within populations (95%), while only 5% of the total variability was among populations (online Supplementary Table S3).
Assessment of genetic diversity based on joint analysis
Clustering based on the joint analysis for phenotypic and molecular marker information formed three groups (Fig. 4). Cluster 1 (red) comprised 21 cultivars. Thirty and six cultivars formed the second cluster (green), with 36% of them from the Department of Collines (Central Benin), and the remaining from Atacora and Donga Departments (North-West). Cluster 3 (blue) comprised of 31 cultivars majority collected from the Donga department and the rest from the other regions.
Fig. 4. Assessment of the genetic diversity among Beninese white Guinea yam accessions through the joint dissimilarity matrix. The colours represent the three clusters: Cluster 1 (red), cluster 2 (green) and cluster 3 (blue).
Discriminant analysis based on phenotypic variables revealed high phenotypic variation among the clusters (online Supplementary Table S4). Cultivars in cluster 1 were characterized by high plant vigour and few spines at the stem's base. The average number of tubers varied from 2 to 3 per plant with a tuber length between 15 and 25 cm and an average yield of 2.18 kg/plant. Cultivars from cluster 2 had vigorous plants, generally no spines on the tuber except Gnidou, and produced long and large tubers (1–2 per plant). Many of the cluster 2 members had high tuber flesh oxidation intensity such as Gnidou, popularly known for this trait. All the cultivars locally known as a ‘Kokoro’ type (intermediate to late maturity cycle) were grouped into cluster 3. Cultivars in cluster 3 were characterized by multiple and small tubers (<15 cm long) per plant that hardly showed the parenchyma's oxidization at peeling.
The distance matrices from the phenotypic and molecular data were weakly correlated (online Supplementary Fig. S6). However, such a relationship was relatively improved with the joint matrix and the respective phenotypic and molecular matrices. The phenotypic distance matrix had a low correlation (0.125) with the joint dissimilarity matrix. In contrast, the correlation between the SNP marker-based Jaccard dissimilarity matrix and the joint matrix was high and positive (r = 0.98) (online Supplementary Fig. S6).
Discussion
Of the 53 phenotypic traits assessed in our study, 30 appeared very informative in discriminating the yam cultivars (online Supplementary Table S2). To save time and resources during yam germplasm characterization, we would recommend focusing on these informative traits in future studies.
The level of diversity in yam cultivars from Benin was found moderate, especially with SNP markers. Our results identified three subgroups of yam cultivars with a low level of admixture. The members of the three sub-groups were generally clustered based on the genetic distance, although those in clusters 1 and 2 were also grouped based on their sex information (Fig. 2) The observed genetic divergence with a low admixture level is due to original differences in domestication and subsequent vegetative propagation by farmers. In Benin, farmers often collect tubers from wild yams, plant them in their fields and select suitable ones through clonal propagation using tubers (Scarcelli et al., Reference Scarcelli, Tostain, Vigouroux, Agbangla, Daïnou and Pham2006; Akakpo et al., Reference Akakpo, Scarcelli, Chaïr, Dansi, Djedatin, Thuillet, Rhoné, François, Alix and Vigouroux2017). This practice is referred to as ‘ennoblement’ and takes 3–6 years until the suitable tuber morphology is achieved. The tradition of raising seedlings from random out-crossing seeds in a farmer's field and further selecting the adapted ones to production niches and uses has not been reported. As a result, the development of new cultivars from cross-breed seeds is a less likely phenomenon for the yam germplasm, even if the varietal mixture of different phenotypes is a common farming practice in many yam growing regions of Africa. Plants of a yam clone in a farmer's field are genetically homogenous with negligible recombination rates, as farmers select their planting material from tubers and not from botanical seed. However, the molecular evidence on ennobled cultivars showed that the tubers collected by farmers from wild environments are often a mixture of wild (D. abyssinica and D. praehensilis) and interspecific hybrid (D. rotundata × wild species) yams (Scarcelli et al., Reference Scarcelli, Tostain, Vigouroux, Agbangla, Daïnou and Pham2006, Reference Scarcelli, Chaïr, Causse, Vesta, Couvreur and Vigouroux2017; Sugihara et al., Reference Sugihara, Darkwa, Yaegashi, Natsume, Shimizu, Abe, Hirabuchi, Ito, Oikawa, Muluneh, Ohta, Matsumoto, Agre, De Koeyer, Pachakkil, Yamanaka, Muranaka, Takagi, White, Asiedu, Innan, Asfaw, Adebola and Terauchi2020). This could partly explain the origin of the genetic admixture identified among the Beninese white Guinea yam accessions. Besides, Loko et al. (Reference Loko, Dansi, Linsoussi, Tamo, Vodouhè, Akoegninou and Sanni2013) showed that farmers often establish yam fields within savannah and forests, a practice that favours gene flows between cultivated yams and their wild relatives.
Our results further demonstrated a high level of gene flow between regions of Benin with no apparent pattern of geographical differentiation in farmer cultivars. This result supports Loko et al.'s (Reference Loko, Ranjana, Agre, Dossou-Aminon, Orobiyi, Djedatin and Dansi2017) findings that used microsatellite markers on Beninese white yam cultivars. Although Tostain et al. (Reference Tostain, Agbangla, Scarcelli, Mariac, Daïnou, Berthaud and Pham2007) agreed on gene flows among regions, they established a geographical pattern between D. rotundata genetic diversity, cropping region and farmers’ crop management practices. However, that geographic pattern hypothesis has not been supported by other diversity studies on Beninese white Guinea yam accessions (Mignouna and Dansi, Reference Mignouna and Dansi2003; Loko et al., Reference Loko, Ranjana, Agre, Dossou-Aminon, Orobiyi, Djedatin and Dansi2017). This gene flow reflects high inter-region seed yam exchange or social and commercial networks for yam in Benin. For instance, farmers often exchange seeds of landraces with other farmers within areas or involve outlining localities, and to some extent, with neighbouring countries such as Nigeria and Togo. They gain access to new landraces, which were adapted to similar environments. Several other studies on cowpea (Fatokun et al., Reference Fatokun, Girma, Abberton, Gedil, Unachukwu, Oyatomi and Boukar2018; Nkhoma et al., Reference Nkhoma, Shimelis, Laing, Shayanowako and Mathew2020), maize (Nelimor et al., Reference Nelimor, Badu-Apraku, Garcia-Oliveira, Tetteh, Agre, N'guetta and Gedil2020) and even yam (Loko et al., Reference Loko, Ranjana, Agre, Dossou-Aminon, Orobiyi, Djedatin and Dansi2017; Adewumi et al., Reference Adewumi, Taah, Asare, Adu and Agre2020, Reference Adewumi, Asare, Adu, Taah, Akaba, Mondo and Agre2021; Bhattacharjee et al., Reference Bhattacharjee, Agre, Bauchet, De Koeyer, Lopez-Montes, Kumar, Abberton, Adebola, Asfaw and Asiedu2020) showed the existence of gene flow among regions and the absence of significant correlations between molecular clustering and geographic origins.
Based on the diversity indexes, the accessions from the department of Collines were more diverse. Traditionally, farmers of this area cultivate multiple yam varieties with different maturity groups such as extra early (5–6 months), early (6–7 months), intermediate (8 months) and late-maturing (9–10 months) to satisfy the steady demand of the international yam market based in Glazoue throughout the year.
Mantel test revealed a low correlation between the dissimilarity matrices originated from the phenotypic and the molecular data. This finding was supported by the membership inconsistency between genomic-based and phenotypic-based clustering. Similar results were reported by several authors using the same approaches on sugar cane, yellow fruit passion, white yam, water yam and cowpea (Sartie et al., Reference Sartie, Asiedu and Franco2012; Alves et al., Reference Alves, Bhering, Rosado, Laviola, Formighieri and Cruz2013; Orgogozo et al., Reference Orgogozo, Morizot and Martin2015; Agre et al., Reference Agre, Asibe, Darkwa, Edemodu, Bauchet, Asiedu and Asfaw2019; Darkwa et al., Reference Darkwa, Agre, Olasanmi, Kohtaro, Matsumoto, Powell, Bauchet, De Koeyer, Muranaka, Adebola, Asiedu, Terauchi and Asfaw2020b; Nkhoma et al., Reference Nkhoma, Shimelis, Laing, Shayanowako and Mathew2020). The low correlation between the phenotypic and the genotypic data could have resulted from the natural and artificial selections on phenotypic variables, as these are under selection and influence of environmental factors. In contrast, the variation detected by molecular markers is commonly non-adaptive, and hence, not subject to natural and or artificial selections (Arnau et al., Reference Arnau, Bhattacharjee, Sheela, Malapa, Lebot, Abraham and Pavis2017). This highlights the importance of combining phenotypic and molecular information while selecting parental lines in breeding programmes to account for their agro-morphological performance and genetic background. Phenotypic traits have the advantage of revealing the agronomic performance of a variety in a particular environment but have limited polymorphism, and they are subjected to changes in environmental conditions (Darkwa et al., Reference Darkwa, Agre, Olasanmi, Kohtaro, Matsumoto, Powell, Bauchet, De Koeyer, Muranaka, Adebola, Asiedu, Terauchi and Asfaw2020b; Nkhoma et al., Reference Nkhoma, Shimelis, Laing, Shayanowako and Mathew2020). Harnessing the advantages of phenotypic and molecular markers improves the grouping of entries in a germplasm collection (da Silva et al., Reference da Silva, Soares, Menezes, Sampaio, da Silva Souza and do Nascimento Macedo2017), which provides a piece of valuable base information for parental selection to realize and sustain genetic gain. The different genetic groups developed through the joint analysis used in this study will be of good use for yam breeding community in West Africa through germplasm exchange. For instance, cultivars with desirable traits such as high yield, tuber quality, disease resistance, etc., have been identified and clustered, and the genetic distance and sex information was provided. This information is critical in selecting parental clones and designing efficient crossing plans. It is noteworthy that more than 90% of the D. rotundata cultivars are diploids (Gatarira, Reference Gatarira2020; Babil et al., Reference Babil, Yamanaka, Girma, Matsumoto, Tamiru-Oli, Bhattacharjee, Abberton, Muranaka, Asiedu, Terauchi and Takagi2021) and the ploidy-level influence on gene flow within this species has not been significant (Mondo et al., Reference Mondo, Agre, Edemodu, Adebola, Asiedu, Akoroda and Asfaw2020, Reference Mondo, Agre, Asiedu, Akoroda and Asfaw2021a, Reference Mondo, Agre, Edemodu, Asiedu, Akoroda and Asfaw2021b). Therefore, cultivars’ agronomic performance, genetic distance and sex information as provided in this study will be a useful asset for yam breeding programmes.
Genetic diversity parameters such as PIC, MAF, observed and expected heterozygosity levels displayed low variability from one region to another, which could be due to the presence of natural gene flow among the study zones and possibly from different neighbouring countries.
Conclusion
This study revealed a low to moderate genetic diversity among different yam-growing areas of Benin as a result of seed exchange among farmers of different regions. However, high genetic diversity was observed within regions due to a range of farmers and other end-users’ preferences within a region. High genetic diversity among plant materials collected within areas as well as information related to their genetic distance and sex provide an opportunity and a good source of selection for plant breeding programmes. We ascertained the relevance of combining molecular markers with phenotypic data for a refined genetic diversity assessment, as it revealed an independent nature of morphological and molecular variations.
Supplementary material
The supplementary material for this article can be found at https://doi.org/10.1017/S1479262121000526
Data
VCF file used in this study including the 88 accessions can be found via this public and open database: https://yambase.org/breeders/trial/796?format=under genotyping data.
Acknowledgements
We acknowledge funding from IITA through a research grant (OPP1052998) received from the Bill and Melinda Gates Foundation and allocated to Benin as a partner to the AfricaYam project. We appreciate the efforts of Mathieu Ayenan and Kwabena Darkwa in critically reviewing this manuscript. We appreciate Beninese farmers for providing studied planting materials and for useful information shared during plant sample collection. Genetic materials were collected under the CBD Nagoya Protocol and were only used for research/breeding purposes.
Author contributions
Alexandre Dansi, Asrat Asfaw, Patrick Adebola and Paterne Agre designed the study; Eric Dadonougbo, Roger Idossou, Anicet Dassou and Anicet Gbaguidi managed the phenotypic data; Paterne Agre managed the molecular data, performed the analysis and wrote the first draft with the contribution from Laura Estelle Loko, Anicet Dassou, Jean Mondo, Yusuf Muyideen, Asrat Asfaw, Robert Asiedu and Alexandre Dansi. Asrat Asfaw critically read and edited the manuscript. All the authors read and approved the content prior to submission.
Conflict of interest
None.
Open access
