Introduction
Yams (Dioscorea spp.) are the world's fourth most economically important edible tuber crop after potatoes, cassava and sweet potatoes (Srivastava et al., Reference Srivastava, Gaiser, Paeth and Ewert2012). They are grown in tropical and sub-tropical regions including West Africa, Asia, Far East, Oceania and tropical America. Tubers of yams are indigenous food and cash crop, predominantly cultivated in the coastal savanna and humid forest agroecologies of western Cameroon, Nigeria, Benin, Togo, Ghana and Cote d'Ivoire. This region, known as the ‘yam belt’, accounts for 96% of the total world yam production (FAOSTAT, 2013). The crop is a dominant source of food and income for up to 60 million smallholder farmers in West Africa, contributing >200 dietary calories per person each day (Degras, Reference Degras1993). Globally, the genus Dioscorea comprises about 450 species (Govaerts et al., Reference Govaerts, Wilkin and Saunders2007). However, only 10 of these species are mainly cultivated for food (Lebot, Reference Lebot2009).
The Genetic Resources Centre (GRC) of the International Institute of Tropical Agriculture (IITA) maintains the largest international collection of yams representing six species, comprising more than 3000 accessions mainly from central and West Africa. These include Dioscorea rotundata Poir., Dioscorea cayenensis Lam., Dioscorea alata L., Dioscorea bulbifera L., Dioscorea dumetorum (Kunth) Pax. and Dioscorea esculenta (Lour.) Burkill. GRC also maintains other wild relatives such as Dioscorea hirtiflora Benth., Dioscorea togoensis R. Knuth, Dioscorea preussi Pax., Dioscorea praehensilis Benth., D ioscorea burkilliana J. Miège and Dioscorea mangenotiana J. Miège. Facilitating the utilization of accessions with desirable traits for cultivar improvement through characterization and classification of collections is one of the user oriented activities of genebanks. This helps in broadening the genetic base of breeders' collection and reduces the challenges of utilizing large collections for evaluation in breeding programs. However, it is always not feasible, mainly for clonally propagated crops, to conduct replicated multi-location characterization/evaluation of such large collections to generate detailed information on every trait of interest of the breeders. This limits the use of germplasm and a large gap still exists between availability and actual use of germplasm collections in breeding programs (Peeters and Galwey, Reference Peeters and Galwey1988; Wright, Reference Wright1997; Upadhyaya et al., Reference Upadhyaya, Dwivedi, Gowda and Singh2007). To overcome these challenges and facilitate easy access to germplasm collections, Frankel and Brown (Reference Frankel, Brown, Holden and Williams1984) introduced the concept of core collection representing about 10% of the entire collection and capturing 70% of the diversity. A core collection of yam was developed by Mahalakshmi et al. (Reference Mahalakshmi, Ng, Obidiegwu, Ogunsola, Lawson and Ortiz2007), consisting of 371 (12.3%) of 3017 accessions representing eight different species collected from 11 West and Central African countries. Recently, additional germplasm have been acquired (453 accessions from Benin and Togo) and information on presence of duplicates and mislabelling was generated through a study on mismatch identification among the entire collection indicating a large discrepancy of about 20% (119 out of 371 accessions were mismatched) (Girma et al., Reference Girma, Korie, Dumet and Franco2012) across individuals within accession. The main objective of this study was therefore to re-visit the existing yam core collection by including information on additional accessions, representing the total diversity with minimum redundancies and mislabelled accessions, in the newly established core collection.
Materials and methods
Plant materials and experimental layout
A total of 4156 individuals representing all accessions found distinct were included in the study. These included 3017 accessions of old collection, 453 of the recently acquired germplasm from Benin and Togo, and 686 individuals from old collection that were found as distinct but previously mislabelled (more details in Girma et al., Reference Girma, Korie, Dumet and Franco2012). The GRC is currently assigning new names for these 686 distinct individuals and the information will be available in the database (http://my.iita.org/accession2/index.jspx). All 4156 individuals were characterized following an augmented design at IITA-Ibadan, Nigeria (Latitude: 7°30′8″N; Longitude: 3°54′38″E). The accessions were planted using a ridge-and-furrow system. Each ridge represented an accession with three plants per ridge and no replication. The distance between two ridges and between two plants within a ridge were 1 m and TDr89/02665, a released variety in Nigeria, was used as check after every 40 accessions. The passport data of entire yam collection showed that most accessions were collected from West and Central Africa (Fig. 1) with the exception of one, D. alata accession, from Tanzania. The list of entire yam collection is available on GRC database and information can be provided on individual accessions on request.
Fig 1. Geographical distribution of accessions of six different Dioscorea spp. maintained at International Institute of Tropical Agriculture (IITA).
Phenotypic traits
A total of 56 morphological traits including above- and below-ground traits (online Supplementary Table S1) were selected from the yam botanical descriptors jointly developed by the International Institute of Tropical Agriculture and International Plant Genetic Resources Institute (IPGRI/IITA, 1997). The morphological traits included in the study were classified as continuous (3), binary (21), ordinal (9) and nominal (23) variables for further data analysis.
Diversity analysis
Shannon–Weaver diversity index for the various traits across the six Dioscorea species were computed based on the following formula (Shannon and Weaver, Reference Shannon and Weaver1949) using MS Excel.
$$H{\rm ^{\prime}} = - \mathop \sum \limits_{i = 1}^s pi\ln (\,pi)$$where; s is the number of phenotypic classes for a character and pi is the portion of the total number of entries belonging to the ith class. The standardized Shannon–Weaver diversity index ranging from 0 to 1 was obtained by dividing H by the log2 of the total number of phenotypic classes.
Selection of the core
Due to the observed differences among different Dioscorea species in respect to the number of informative traits, the statistical analysis was carried out independently for each species for establishing the core collection and results were then combined at the end to form a unique core subset representing about 20% of the entire collection. The steps for analysis, as proposed by Franco et al. (Reference Franco, Crossa and Deshpande2010) were: (1) defining the informative variables; (2) using the Hierarchical Multiple Factor Analysis (HMFA) (Le Dien and Pagés, Reference Le Dien and Pagés2003) to combine different kinds of variables (ordinal, nominal, binomial and continuous) in an equilibrated way; (3) obtaining the Principal Axes (PA) coordinates from the HMFA explaining 90% or more of the total inertia (the measure of variability for particular kind of variables) for each accession; (4) classifying the accessions into clusters by the Mixture of Normal Distributions method applied on the mentioned coordinates; (5) assigning and identifying appropriate number of accessions from each cluster using the D-method (Franco et al., Reference Franco, Crossa, Taba and Shands2005), where the assigned number is proportional to the cluster diversity, measured by the average value of the Gower Distance (Gower, Reference Gower, Armitage and Colton2005) between pair of accessions belonging to each cluster; (6) re-sampling or bootstrapping process to select 1000 independent ‘candidate’ core subsets (each one consisting of 843 accessions) by a stratified random sampling procedure and obtaining the average of distances for each of them; (7) the candidate core subset showing the highest average Gower distance, that is, the maximum of diversity, was selected as the final core subset. R (R Core Team, 2011), packages FactoMineR (Josse and Husson, Reference Josse and Husson2008) and Cluster (Kaufman and Rousseeuw, Reference Kaufman and Rousseeuw1990) were used for the HMFA analysis and to calculate Gower Distance, respectively.
Principal component analysis (PCA) was further computed to visualize the representativeness of the core collection with the entire collection for each Dioscorea spp. using SAS software version 9.4 (SAS Institute Inc., 2013). The method of linear transformation, maximum total variance of PRINQUAL, was adopted using SAS procedure PRINQUAL for qualitative dataset. The procedure produced a set of transformed variables in form of scores, which was then used in a PCA, and the first two dimensions used to generate the plots. Furthermore, Shannon–Weaver diversity indices estimated for the entire and the core collection was compared.
Results
The core subset
The procedure used in the present study to establish the revised yam core collection resulted in the selection of 843 accessions (20.3% of the entire collection) from the international collection maintained at IITA (Table 1; Fig. 1). The core reflected the predominance of D. rotundata (620 accessions; 73.54%) followed by D. alata (180 accessions; 21.35%), D. bulbifera (14 accessions; 1.66%), D. cayenensis (12 accessions; 1.42%), D. dumetorum (12 accessions; 1.42%) and D. esculenta (5 accessions; 0.59%) (Table 1). In terms of countries, Togo, Nigeria and Benin represented the largest collection in descending order for both entire and core collection representing all the six species.
Table 1. List of Dioscorea species with respective number of accessions and country of origin in the entire and core collection (within bracket)
Selection of accessions to establish the core subset
Among the morphological descriptors used to re-define the core collection, not all were informative and varied across species. Of the total 56 traits only 43 in D. alata, 23 in D. bulbifera, 31 in D. cayenensis, 37 in D. dumetorum, 16 in D. esculenta and 48 in D. rotundata were informative (online Supplementary Table S1). The methodology used for selecting accessions was proportional to the average of the Gower distances between accessions for each species included in the study (Table 2). Interestingly, the groups for D. rotundata and D. alata, representing the larger proportion of the entire collection, showed relatively lower distances between accessions, while the groups for D. bulbifera, D. dumetorum and D. esculenta with fewer accessions showed higher distances between accessions (Table 2). This indicated the importance of using HMFA method in selecting accessions from each group representing the distances between accessions within each group and not size of the groups. The proportionally lower values of distances within larger groups may be due to lower values of diversity within those groups, since larger groups may have presence of redundant accessions or related accessions with very low distances between them.
Table 2. Distribution of accessions in different species groups and phenotypic diversity based on Gower's distances
G-dist, Gower's distances.
Phenotypic diversity
The informative morphological traits (continuous, ordinal, nominal and binary) per species were transformed into principal axes coordinates to explain the total variation. The comparison of effects of different types of variables (continuous and ordinal versus nominal) using the first 50 principal axes that explained about 90% of the total variation, showed a 50:50% contributions for each of these variables (data not shown). Further analysis indicated that all of the continuous and ordinal variables, 18 out of 20 binary variables and 22 out of 23 nominal variables contributed significantly (P < 0.05) towards grouping of accessions into different clusters (data not shown).
The Shannon–Weaver diversity (H′) estimated for the entire and core collection for each trait across different Dioscorea spp. is presented in online Supplementary Table S2. The index is commonly used in estimating phenotypic diversity to estimate allelic richness and evenness. A low index for a particular trait indicates low genetic diversity. The results showed that there were no significant differences in average H′ index for the entire and revised core collection across all the traits (online Supplementary Table S2). This is also the case across each individual trait wherein no significant differences were observed between the entire and the core collection. The core retained the overall phenotypic diversity for each trait and is representative of the entire collection across all the six species (online Supplementary Table S2). In some cases, such as D. alata, the core collection showed a gain of about 10–12% in the average distances between accessions. Similarly, PCA graphs generated based on morphological traits for entire and core collection and principal axes for each Dioscorea spp. showed the representativeness and spread of the accessions in the core and entire collection (Fig. 2). For all the species, the number of accessions selected represented about 20% of the total collection and was proportional to the size of each group. Overall, the revised core reflected the predominance of accessions of D. rotundata and D. alata, which is in accordance with the entire collection (Table 1).
Fig 2. Principal component analysis representing the main and core collection of six different Dioscorea spp. maintained at International Institute of Tropical Agriculture (IITA).
Flowering is a very important breeding requirement in any crop and most of the important yam species cultivated for their edible tubers do not flower. In cases where plants flower, the male-to-female ratio is very high. In the present study, data on flowering showed the same trend in D. rotundata while the female-to-male ratio was high for D. bulbifera (Table 3). There were D. rotundata accessions with monoecious plants (predominant male flowers with female flowers in the same plant and vice-versa) and many which showed no flowering. Most of D. bulbifera, D. alata and D. esculenta accessions did not produce any flowering as well. The accessions of D. cayenensis produced male flowers while that of D. dumetorum produced male and female flowers separately with many that did not produce any flowers (Table 3).
Table 3. Flowering data on different Dioscorea spp.
Yams are polyploid in nature with ploidy levels of 2x (diploid), 3x (triploid) and higher ploidy levels. Flow cytometer is used to determine ploidy based on relative fluorescence intensity of a reference (standard). In the current study, the ploidy status of only D. rotundata accessions could be determined. For other species, either the standard was not available or has been lost over the years. Among D. rotundata accessions, 88.6% were diploid (2n = 40) and remaining 11.4% were triploid (2n = 60) (Babil et al., unpublished; data not shown).
The list of yam accessions in the core collection with details on species, passport and characterization data are available on request and uploaded at http://www.iita.org/genetic-resources-center.
Discussion
In the present study, a mixture of different types of measured variables was used to group the accessions. Additionally, the accessions were classified into different groups of different sizes based on species and geographical origin. This posed a challenge in grouping the accessions representing maximum diversity present in the entire collection. This was addressed by combining different statistical methods: the HMFA allowed the mixture of different types of variables and their effective contribution in the classification of accessions in different groups; the D-method for the selection of accessions from each group/cluster proportional to the within-cluster diversity; and 1000 iterations to select candidate core collection through independent stratified random sampling processes (Franco et al., Reference Franco, Crossa and Deshpande2010). This allowed the constitution of the final core collection that not only represented the mean of each of the variables but also maintained the phenotypic diversity of the entire collection. Similar approach of using different types of measured variables and statistical method was used to define the cassava core collection (Bhattacharjee et al., Reference Bhattacharjee, Dumet, Ilona, Folarin and Franco2012).
A core collection was previously developed for yams by Mahalakshmi et al. (Reference Mahalakshmi, Ng, Obidiegwu, Ogunsola, Lawson and Ortiz2007). However, a later study showed significant percentage of mislabelling and mixture of tubers within accessions (Girma et al., Reference Girma, Korie, Dumet and Franco2012). Accession mismatch/mislabelling is a common problem with most of the clonally propagated crops and could generate from serious mix-ups during planting or storage and also due to multiple collection missions, during which different names might have been given to the same clone with yam being maintained as multiclonal crop. Additionally, yam tubers are cut into mini-setts for planting and there are possibilities of mix-ups (human error) between accessions while planting several mini-sets per accession for regeneration or characterization. The study by Girma et al. (Reference Girma, Korie, Dumet and Franco2012) therefore identified several mislabelled accessions that were unique and have been considered as individual accessions. The GRC has generated unique identification numbers for these newly identified individuals or accessions. Moreover, additional yam germplasm were acquired at the IITA genebank. Considering that core collections are dynamic in nature, an attempt was therefore made to include the information on newly identified accessions and additionally acquired accessions to revise the yam core collection. The re-defined core collection consisted of 843 accessions representing about 20% of entire collection. Generally, a core collection should represent 10% of the collection (Frankel and Brown, Reference Frankel, Brown, Holden and Williams1984), however, a recent study on determining the efficiency and effectiveness of sampling strategies on development of core collections used five different fractions (10, 15, 20, 25 and 30%) (Studnicki et al., Reference Studnicki, Madry and Kociuba2010). They suggested that increasing sample fraction improved representativeness of phenotypic diversity of the core collections and the core collection representing at least 20% of the entire collection is sufficient to provide enough information. They also suggested that there were no significant differences between the core collections developed using 20% fraction and 25 or 30% fractions (Studnicki et al., Reference Studnicki, Madry and Kociuba2010).
The newly established yam core collection represented 843 accessions of six Dioscorea spp. and 11 countries with two D. cayenensis accessions whose geographical origin was unknown. Some of the countries such as Guinea Bissau and Tanzania were not represented in the core collection, as there was only one accession from each of these countries in the entire collection. The majority of the accessions in the core collection belonged to only two Dioscorea spp. including D. rotundata and D. alata representing 73.54 and 21.35% of the entire collection, respectively. There are an estimated 450–600 species in the genus Dioscorea with worldwide geographical distribution (Govaerts et al., Reference Govaerts, Wilkin and Saunders2007; Considine and Considine, Reference Considine and Considine2012). However, the international Dioscorea germplasm collection maintained at IITA genebank represents only six species and are mostly from West Africa, largest being D. rotundata with a good number of accessions from Togo, Nigeria and Benin. In general, the entire yam germplasm collection at IITA represents a wide gap in terms of species and geographical representation. The most obvious reason for such limited representation could be due to large number of Dioscorea spp. (450–600) with wide geographical distribution making it difficult to have enough funding for collection missions. IITA's primary focus is on Africa and more specifically in West Africa, where D. rotundata is considered as the most important species, hence large representation of D. rotundata in the collection. The unequal representation of even West African countries for D. rotundata collection can be linked with the limited opportunities for collection or acquiring of germplasm from different countries in the past decades. Efforts are underway to widen the international Dioscorea (yam) collection maintained at IITA genebank through collection missions and liaising with different countries, mainly within West and Central Africa, which resulted in the inclusion of additional accessions and the need to re-define the core collection.
The revised core collection therefore will serve as a reliable reference for establishment of further subset or mini core collection in addition to creating an entry point for the efficient evaluation and further utilization of the yam germplasm in breeding programs. Most importantly, the accessions in the revised core collection are true-to-type representing higher phenotypic diversity for accessions in some of the species compared with the entire collection. The revised core will also provide guidance to the genebank managers while acquiring new accessions in the collection. The accessions in the core being true-to-type, can be used for large-scale multi-location phenotypic characterization and molecular profiling to identify genetically diverse parents for making crosses to generate segregating populations for target traits and also broadening the genetic base of the accessions.
Yams show variation in ploidy status within and between species. Although differences in ploidy levels do not reflect any unusual morphological features in the yam plants but affects efficient hybridization schemes in yam breeding and improvement programs (Dansi et al., Reference Dansi, Pillay, Mignouna, Mondeil and Dainou2001a, Reference Dansi, Mignouna, Pillay and Zokb). Efforts are underway at IITA in collaboration with Institute of Experimental Botany, Czech Republic, to determine the ploidy status of reference clones of all the important Dioscorea spp. using conventional karyotyping. This will facilitate rapid and easy ploidy determination of all the accessions in the core collection using high-resolution flow cytometry. The current yam breeding and improvement work in West and Central Africa is concentrated mainly on D. rotundata and D. alata, and information on flowering and ploidy on core accessions will help in initiating intra- and inter-specific hybridizations. Currently, efforts are underway at IITA using biotechnology tools to generate successful hybridizations among different species such as D. rotundata with D. alata, D. bulbifera, D. esculenta and D. dumetorum. The establishment of core collection makes it easy to select parents/lines from each species for use in the breeding program.
Supplementary material
The supplementary material for this article can be found at https://doi.org/10.1017/S1479262117000144.
Acknowledgements
The authors like to thank all the field technicians of Genetic Resources and yam breeding units, IITA involved in morphological characterization. The first author, Gezahegn Girma, acknowledges support the Dutch ministry of foreign affairs/APO programme. All other co-authors would like to thank the Global Crop Diversity Trust for funding towards long-term conservation and maintenance of the core collection.
