1. Introduction
Considerable confusion in the synonymy and classification of various Vigna species exists in the literature (Verdcourt, Reference Verdcourt1970; Fery, Reference Fery and Janick1980). The genus Vigna comprises seven subgenera and more than 80 species. Some of the species adapt well to a wide range of environmental conditions (such as poor soils and drought) and have been domesticated to cultivated species (Faris, Reference Faris1965; Verdcourt, Reference Verdcourt1970; Santalla et al., Reference Santalla, Power and Davey1998). Cultivated Vigna species are an important protein source in countries where people have limited access to food rich in protein (Singh, Reference Singh, Singh & and Jauhar2005). The main cultivated species worldwide include the following: five Asian beans: moth bean (Vigna aconitifolia (Jacq.) Marechal), azuki bean (Vigna angularis (Willd.) Ohwi and Ohashi), black gram (Vigna mungo L.), mung bean (Vigna radiata L.), rice bean (Vigna umbellata Thunb.); two African beans: bambara groundnut (Vigna subterranea L.) and cowpea (Vigna unguiculata (L.) Walp); and American Vigna beans (Jaaska & Jaaska, Reference Jaaska and Jaaska1990; Jaaska, Reference Jaaska1999, Reference Jaaska2001).
Isoenzymes as biochemical markers have been used to assess genetic diversity and reveal phylogenetic relationships among Vigna species (Jaaska & Jaaska, Reference Jaaska and Jaaska1990; Pasquet, Reference Pasquet1999, Reference Pasquet2000). The phylogenetic relationships and genetic diversity among and within Vigna species were first assessed by restriction fragment length polymorphism (RFLP) DNA markers (Fatokun et al., Reference Fatokun, Danesh, Young and Stewart1993), and then by random amplified polymorphic DNA (RAPD) (Kaga et al., Reference Kaga, Tomooka, Egawa, Kosaka and Kamijima1996; Santalla et al., Reference Santalla, Power and Davey1998; Lakhanpaul et al., Reference Lakhanpaul, Chadha and Bhat2000; Mimura et al., Reference Mimura, Yasuda and Yamaguchi2000; Xu et al., Reference Xu, Tomooka, Vaughan and Doi2000; Amadou et al., Reference Amadou, Bebeli and Kaltsikes2001; Ba et al., Reference Ba, Pasquet and Gepts2004; Diouf & Hilu, Reference Diouf and Hilu2005) and amplified fragment length polymorphism (AFLP) (Tomooka et al., Reference Tomooka, Yoon, Doi, Kaga and Vaughan2002; Zong et al., Reference Zong, Kaga, Tomooka, Wang, Han and Vaughan2003; Seehalak et al., Reference Seehalak, Tomooka, Waranyuwat, Thipyapong, Laosuwan, Kaga and Vaughan2006; Yoon et al., Reference Yoon, Lee, Kim and Baek2007) DNA markers. As more DNA sequence information is now available, internal transcribed spacer (ITS) sequences (Doi et al., Reference Doi, Kaga, Tomooka and Vaughan2002; Goel et al., Reference Goel, Raina and Ogihara2002), simple sequence repeat (SSR) markers (Li et al., Reference Li, Fatokun, Ubi, Singh and Scoles2001; Kumar et al., Reference Kumar, Tan, Quah and Yusoff2002; Wang et al., Reference Wang, Kaga, Tomooka and Vaughan2004; Gillaspie et al., Reference Gillaspie, Hopkins and Dean2005) and DNA amplification fingerprinting (DAF) (Simon et al., Reference Simon, Benko-Iseppon, Resende, Winter and Kahl2007) were used for assessing the phylogenetic relationships and genetic diversity. Although various DNA markers have been developed from different Vigna species, there are neither a common nor a sufficient set of robust DNA markers available for evaluation of germplasm applicable to all Vigna species. Recently, gene-derived markers were developed across 15 legumes (Choi et al., Reference Choi, Luckow, Doyle and Cook2006). Since these markers are derived from putative genes, they may be a good source to reveal phylogenetic relationships and assess genetic diversity among and within Vigna species.
The US germplasm resource for Vigna species is maintained at the USDA-ARS Plant Genetic Resources Conservation Unit (PGRCU) located at Griffin, GA, USA. The phylogenetic relationships in the USDA Vigna germplasm collection are unclear and their genetic diversity is unknown. Revealing the phylogenetic relationships and assessing genetic diversity will help develop strategies for better organization and management of existing as well as further acquisitions of Vigna germplasm. The objectives of the present study were to: (i) reveal phylogenetic relationships and assess genetic diversity of Vigna species in the USDA collection using gene-derived DNA markers, (ii) sequence amplicons generated from gene-derived primers to identify polymorphisms and (iii) evaluate the effectiveness of gene-derived markers in revealing the phylogenetic relationships and assessing genetic diversity among and within Vigna species.
2. Materials and methods
(i) Plant materials and DNA extraction
Taxonomic classifications of accessions used in the present study are based on the Germplasm Resources Information Network (GRIN; found at http://www.ars-grin.gov/npgs/index.html). Forty-eight accessions from several Vigna species (Table 1) were used in this experiment and all accessions were diploids containing 11 pairs of chromosomes (2n=2x=22) (Singh, Reference Singh, Singh & and Jauhar2005). Among them, 12 accessions were from V. unguiculata L. (cowpea, representing six subspecies); five accessions from V. angularis Willd. (azuki bean); four accessions each from V. radiata L. (mung bean), V. mungo L. (black gram), V. umbellata Thunb. (rice bean) and Vigna oblongifolia A. Rich.; three accessions each from V. aconitifolia Jacq. (moth bean) and V. subterranea L. (bambara groundnut); two accessions each from Vigna adenantha G. Mey., Vigna caracalla L. and V. luteola Jacq.; and one accession each from Vigna longifolia Verdc. and Vigna vexillata L. One accession from Phaseolus vulgaris L. (common bean), which is closely related to the Vigna genus (Verdcourt, Reference Verdcourt1970), was also included as an outgroup in the present study. Leaf tissue samples were collected from plants grown in a greenhouse at Griffin, GA, USA. DNA was extracted from leaf tissue using an E.Z.N.A. Plant DNA Miniprep kit from Omega Bio-Tek (Doraville, GA, USA). The DNA was then diluted to 10 ng/μl and later used as a template for PCR.
Table 1. Selected accessions from Vigna species
Letters in parentheses: m, mottle; sol, solid; spec, speckled; str, streaked.
(ii) PCR and PCR product separation
Thirty pairs of primers tested in mung bean (V. radiata) were selected from GenBank® (BV164338 to BV165946) based on published information (Choi et al., Reference Choi, Luckow, Doyle and Cook2006) and are listed in Table 2. All PCR reactions, programmes and product separations on agarose gels were performed by following the method described previously by Wang et al. (Reference Wang, Mosjidis, Morris, Dean, Jenkins and Pederson2006).
Table 2. Selected primers
a Markers were monomorphic for Vigna germplasm.
IS, intron-spanning marker; ED, exon-derived marker; STS, sequence-tagged site-derived marker.
(iii) Allele sequencing
Before sequencing, PCR products were checked on a 3% agarose gel to verify that only a single band was produced from each sample. The PCR product was treated with 1 μl of exonuclease I (10 units/μl) and 1 μl of shrimp alkaline phosphatase (1 unit/μl) (GE Healthcare, Piscataway, NJ, USA) for every 12 μl of PCR reaction to digest single-stranded DNA and cleave the 5′-phosphate, respectively. The PCR product was also cleaned with a Qiagen PCR cleanup kit (Qiagen, Valencia, CA, USA) to remove excess nucleotides, primers, enzymes and other impurities. Then, 1 μl of the cleaned product was run on an agarose gel with a quantitative marker (Invitrogen, Carlsbad, CA, USA) to determine product concentration and thus prepare the sample for sequencing. Sequencing reactions were prepared by following the instructions of the DTCS Quick Start sequencing kit (Beckman Coulter, Fullerton, CA, USA). The sample was sequenced bi-directionally and pUC18 was also sequenced as a positive control. Each sample was sequenced twice to verify the fidelity of the sequenced bases. Samples were injected and sequenced on a Beckman CEQ 8000 by using the LFR-1 method. The sequence module of the software package CEQ 8000 Genetic Analysis System version 8.0.52 from Beckman was used to call the bases after the sequencing was performed. The forward and reverse strands were edited and aligned using AlignIR version 2.0 (LI-COR, Lincoln, NE, USA).
(iv) Morphology comparison
Seeds were harvested from plants grown in a greenhouse at Griffin, GA, USA. Seeds mainly representing the accession were weighed. Seed-coat colour and size were scanned and recorded with a Hewlett-Packard Scanjet 7400C. In order to confirm the classification, more accessions from the same species or subspecies were requested from the Griffin seed store and compared regarding their seed morphology.
(v) Data analysis
Strong clear bands on the gel images were scored as either present (1) or absent (0) for DNA fragment analysis. The data were entered into a binary matrix for analysis. A distance matrix was created between all pairwise combinations by using the proportion of shared allele algorithm in the program MICROSAT v.1.5 (Minch et al., Reference Minch, Ruiz-Linares, Goldstein, Feldman, Kidd and Cavalli-Sforza1997). One hundred bootstrap replicates were generated and a neighbour-joining tree and a consensus tree were constructed using PHYLIP v.3.6 (Felsenstein, Reference Felsenstein2005). The trees were then viewed and printed using TreeView (Page, Reference Page1996).
Thirty-two consensus sequences were imported into ClustalX (Thompson et al., Reference Thompson, Gibson, Plewniak, Jeanmougin and Higgins1997) and aligned for DNA sequencing data analysis. Low gap penalties (gap penalty=10 and gap extension=0·1) were applied using the slow and accurate pairwise alignment of ClustalX. The resulting alignment was evaluated for maximum pairwise identity. In general, high gap penalties are suitable for intraspecific data, whereas low gap penalties are suitable for interspecific data. The sequence data were imported into Phylip v.3.6 (Felsenstein, Reference Felsenstein2005) and maximum likelihood (DNAML) was employed. The program SEQBOOT from the PHYLIP package was used to perform bootstrapping with 100 replicates to test the stability of the clades. CONSENSE was used to create consensus trees from bootstrap replicates.
3. Results and discussion
(i) Gene-derived markers, genetic diversity and phylogenetic relationships
All 30 gene-derived primers amplified DNA fragment(s) from the accessions tested. Employed as DNA markers, 26 (86·7%) were polymorphic and 4 (13·3%) were monomorphic. There were no polymorphic markers identified within an accession. All of the markers derived from the intron-spanning (IS) regions were polymorphic, whereas only four of the seven markers from the exon-derived (ED) regions were polymorphic (Table 2). For example, marker BV165019 derived from the intron-spanning region of the gene cationic peroxidase 2 (COPX2) revealed interspecific polymorphisms but not intraspecific polymorphisms, whereas the marker BV 165028 from the exon-derived region of carboxyl-terminal peptidase (CTP) was monomorphic for all the accessions examined (Fig. 1). The results from gene-derived markers demonstrated that at the DNA fragment level, (i) intron-spanning markers were more effective in revealing polymorphisms than exon-derived markers and (ii) gene-derived markers were more effective in revealing polymorphisms among species than within species (Fig. 1 and Table 2). However, future studies should include more accessions from a single species to evaluate intraspecific polymorphism.
Fig. 1. Amplicons generated by PCR and separated by electrophoresis. Each well contains either 10 μl of molecular marker (100 bp ladder, 250 ng) or 12·5 μl of PCR products. The fragments were separated by electrophoresis on a 3% agarose gel. PCR products were generated with the primer pairs of BV165019 and BV165028 from Vigna radiata.
A total of 134 polymorphic bands (DNA fragments) were observed with an average of approximately five bands per marker, which were subsequently used for phylogenetic analysis. A dendrogram generated from gene-derived DNA marker data is shown in Fig. 2 and the accessions examined clustered into four main groups (American bean group, African bean group, Asian bean group and intermediate group). One accession (Plant Introduction number (PI) 633451) from P. vulgaris, which was used as the outgroup, was different from most of the accessions, but most similar (genetic distance of 0·027) to one accession (PI 146800) from V. caracalla. This was not surprising because P. vulgaris was taxonomically separated from the Vigna genus only since 1970 (Verdcourt, Reference Verdcourt1970).
Fig. 2. Neighbour-joining tree of Vigna species. Bootstrapping was performed with 100 replicates and values greater than 40% were placed on the branches.
Group I included three accessions: two accessions (PI 312898 and PI 319448) from V. adenantha, and one accession (PI 322588) from V. caracalla. Species of V. adenantha and V. caracalla belong to the same subgenus Sigmoidotropis (Piper) and were also classified as American beans using biochemical markers (Jaaska, Reference Jaaska2001). Interestingly, PI 312898 was closely related (distance=0·013) to PI 322588, which was from a different species (V. caracalla, supported by a 59% bootstrap value). Species within group I was closely related to P. vulgaris. This result was similar to previous phylogenetic analyses of the Phaseolus–Vigna complex (Jaaska, Reference Jaaska2001; Goel et al., Reference Goel, Raina and Ogihara2002).
Group II contained only one accession (PI 310294), which belongs to V. longifolia. This group is designated as an intermediate group.
Group III contained 20 accessions (from PI 164419 to PI 171435 in Fig. 2) covering five species or five Asian beans, representing five subgroups. Accessions from each species formed a cluster. Three accessions classified as V. aconitifolia (moth bean) were examined. Two accessions (PI 165479 and PI 372355) clustered together with little genetic variation (genetic distance=0·001) and were supported by a 66% bootstrap value. These accessions were different (genetic distance=0·027) from accession PI 164419. Therefore, some genetic diversity was detected within moth bean. Four accessions of V. mungo var. mungo (black gram, from PI 208462 to PI 305073) were examined and they formed a distinct cluster. Five accessions of V. angularis (azuki bean, from PI 527686 to PI 360707) were examined and very little genetic difference was detected among these accessions as demonstrated by the short branches and small genetic distance values among these accessions ranging from 0·0008 to 0·0018. Furthermore, strong support for monophyly was apparent with the support of a 73% bootstrap value. The accession, PI 527686 was previously determined to be the wild form (or progenitor) of cultivated azuki bean (Mimura et al., Reference Mimura, Yasuda and Yamaguchi2000) and the phylogeny supports this notion. Four accessions were examined within V. umbellata (rice beans). Accessions (PI 247689 and PI 275636) and (PI 173933 and PI 208460) formed two small clusters that were supported by 71 and 62% bootstrap values, respectively. Four accessions of V. radiata (mung bean, PI 171435, PI 164301, PI 381351 and PI 427064) were examined and formed a small cluster. Very limited genetic diversity was detected in mung bean (genetic distance ranging from 0·0009 to 0·003). This result was consistent with the genetic diversity revealed by SSR markers in mung bean (Gillaspie et al., unpublished results) and EcoTILLING (Barkley et al., Reference Barkley, Wang, Gillaspie, Dean, Pederson and Jenkins2008), which may suggest a narrow genetic base for V. radiata. Low levels of genetic diversity were also revealed among 32 Indian mung bean cultivars (Lakhanpaul et al., Reference Lakhanpaul, Chadha and Bhat2000). Based on branch nodes from our results, mung bean (V. radiata) was genetically very different from the other four beans.
Group IV contained 22 accessions covering five species (including two African beans: bambara groundnut and cowpea), representing five subgroups. Accessions from each species formed a cluster. The species within group IV originated in Africa and therefore this group was named the African bean group. Two accessions (PI 292866 and PI 406329) from V. luteola clustered with a genetic distance of 0·008. Four accessions were examined within V. oblongifolia and they formed a small cluster with the support of a 72% bootstrap value. However, PI 406358 was genetically similar to three other accessions with genetic distances ranging from 0·011 to 0·014 (PI 292868, PI 181585 and PI 292872; Fig. 2). Based on phylogenetic analysis, it seems that the V. luteola species was closely related (distance=0·22–0·31) to V. oblongifolia with the support of a 60% bootstrap value. One accession (PI 406390) from V. vexillata clustered closely to V. unguiculata and V. subterranea, which was consistent with the early RFLP analysis (Fatokun et al., Reference Fatokun, Danesh, Young and Stewart1993) and morphological observations (Bisht et al., Reference Bisht, Bhat, Lakanpaul, Latha, Jayan, Biswas and Singh2005). Three accessions (PI 378867, PI 240867 and PI 245951) were examined within V. subterranea (bambara groundnut) and they formed a small cluster with the support of a 67% bootstrap value. However, PI 240867 and PI 245951 were closely related genetically (distance=0·005) with the support of a 69% bootstrap value and were distinct from PI 378867. In a previous study, genetic classification of V. subterranea accessions was related to geographic origin and accessions collected from Nigeria were very different from accessions collected from Zimbabwe (Amadou et al., Reference Amadou, Bebeli and Kaltsikes2001). In the present study, PI 378867 collected from Nigeria was also very different from PI 245951 (genetic distance of 0·01) collected from Zimbabwe. Twelve accessions (PI 632904–PI 582578) from cowpea (V. unguiculata) clustered together, forming a subgroup. There were six subspecies examined within V. unguiculata (Table 1). The accession (PI 632910) from subspecies pubescens may be closely related (distance=0·005–0·007) to the accessions (PI 632903 and PI 632904) from subspecies stenophylla and formed a small cluster. Four accessions (PI 582578, PI 582469, PI 582470 and PI 612607) from subspecies unguiculata, three accessions (PI 610582, PI 419163 and PI 215659) from subspecies sesquipedalis, one accession (PI 292883) from subspecies dekindtiana and one accession (PI 291384) from subspecies cylindrica formed a small cluster with the support of a 68% bootstrap value. Obviously, some genetic diversity was detected among these subspecies but compared with other species examined within group IV; the genetic diversity detected was minimal, as demonstrated by short branch lengths and low genetic distance among accessions ranging from 0·005 to 0·013 within V. unguiculata. This result was consistent with results from other studies (Li et al., Reference Li, Fatokun, Ubi, Singh and Scoles2001; Diouf & Hilu, Reference Diouf and Hilu2005) in which a narrow genetic base was also found in cowpea breeding lines and local varieties from Senegal. A single domestication event between wild and cultivated cowpea may be the explanation for the narrow genetic base within cowpea (Coulibaly et al., Reference Coulibaly, Pasquet, Papa and Gepts2002; Ba et al., Reference Ba, Pasquet and Gepts2004). Another possible explanation is that the type of DNA markers employed may also affect the level of polymorphism revealed. To reveal genetic diversity of cowpea (V. unguiculata var. unguiculata), SSR, RAPD and gene-derived markers were used in two previous studies as well as in the present study, respectively. When 26 DAF primers were employed, 54 cowpea accessions (V. unguiculata var. unguiculata) were classified into separate groups (Simon et al., Reference Simon, Benko-Iseppon, Resende, Winter and Kahl2007). DAF may be a highly efficient system for the generation of polymorphic DNA markers for revealing cowpea genetic diversity.
(ii) DNA sequencing, genetic diversity and phylogenetic relationships
To detect polymorphism at the DNA sequence level, 32 amplicons generated from the marker BV165019 were sequenced and the sequence alignment is shown in Fig. 3. Possible sequence errors were identified at a primer site when the consensus sequences were constructed from forward and reverse reads. The first few base pairs of the sequences with possible errors were removed when the phylogenetic tree was generated from sequence data. The phylogenetic tree generated from sequence data is shown in Fig. 4. Comparing the sequence alignment of the common bean (P. vulgaris, PI 633451) with the remaining sequences derived from species within the genus Vigna, several small deletions or insertions (indels) were identified. This implies that indels may play an important role in differentiation and speciation (Fig. 3). According to the phylogenetic analysis from DNA bands (size of the DNA fragments), PI 322588 from V. caracalla formed a cluster with accessions (PI 312898 and PI 319448) from V. adenantha. Our sequence data further confirmed the above phylogenetic relationship. Within the species of V. adenantha, there was a base pair deletion at position 340 observed between PI 319448 and PI 312898 (Fig. 3).
Fig. 3. Sequence alignment of gene-derived marker BV165019 alleles generated from selected Vigna accessions produced by using AlignIR version 2.0.
Fig. 4. Maximum likelihood tree of Vigna species generated from BV165019 amplicons. Bootstrapping was performed with 100 replicates and values greater than 40% were placed on the branches.
Although accessions from different subspecies within cowpea were sequenced, there were no sequence polymorphisms identified (Fig. 3). The accessions from cowpea formed a cluster with a bootstrap value of 57% and no variation among accessions. The low genetic diversity revealed within cowpea may be explained by a single domestication event (Pasquet, Reference Pasquet1999). In comparison with cowpea accessions, V. vexillata (PI 406390) had a four-base-pair insertion (154–157), a large deletion (182–194) and one-base-pair deletion (259) (Fig. 3). Within the Asian bean group, two accessions from each species were sequenced and all sequenced Asian bean accessions formed a group (Fig. 4) with little diversity among the species within this group. Sequence variation (point mutation, insertion or deletion) was identified among these five species. However, only one point mutation (A/G) at position 196 was identified within the species V. mungo (Fig. 3). The species V. longifolia (PI 310294) within the African bean group was distinguished from other African bean accessions by several smaller deletions (Fig. 3). The species V. subterranea (PI 240867 and PI 245951) were separated from the species V. luteola and V. oblongifolia. There was a point mutation identified (from T to C) at position 258 within the species V. luteola. Overall, sequencing gene-derived amplicons detected more variation among species than within species, which was consistent with the results from detecting size difference of DNA fragments on an agarose gel. The topologies of the two dendrograms generated from DNA fragment data and DNA sequence data were very similar (Figs 2 and 4). The method of sequencing DNA amplicons for characterization of germplasm may be expensive, but the method for separating DNA fragments on agarose gels may require processing a large number of DNA markers to be effectively utilized in some genera.
(iii) Seed morphological observation and phylogenetic relationships
Seed morphology from 48 accessions was observed and recorded (shown in Fig. 5). The phylogenetic analysis showed that two accessions (PI 146800 and PI 322588) from V. caracalla were placed into different groups. Accession PI 146800 had small speckled tan colour seeds (1·93 g per 100 seeds), whereas PI 322588 had large solid brown seeds (5·22 g per 100 seeds) (Fig. 5 and Table 1). Furthermore, within 313 base pairs of the sequenced amplicons, eight point mutations were identified between these two accessions (Fig. 3). Given all the collected data, it is possible that these two accessions may belong to different species. Actually, the seed morphologies of PI 322588 (V. caracalla) and PI 312898 (V. adenantha) were very similar (Fig. 5) and these two accessions were also clustered into the same subgroup from the phylogenetic analysis of markers and sequence data (Figs 2 and 4). It is suspected that these two accessions may belong to the same species (V. adenantha). To confirm this speculation, more accessions classified as V. adenantha were requested from the Griffin seed store and their seed morphology was compared. The seed morphology of these two accessions and other accessions from the same species was very similar. Further confirmation of the possible misidentified accessions will include collecting more observational data (for example, seedling morphology, flowering characteristics and other traits) and experiments need to be conducted. Future work will include growing these accessions and collecting descriptor data to determine whether this accession has been misclassified or mislabelled during curation of this crop.
Fig. 5. Seed-coat colours from one common bean accession (Phaseolus vulgaris L.) and 47 Vigna accessions. From left to right: the eight accessions of the first row are PI 633451 (P. vulgaris L.), PI 164419 (Vigna aconitifolia Jacq.), PI 165479 (V. aconitifolia Jacq.), PI 372355 (V. aconitifolia Jacq.), PI 312898 (Vigna adenantha G. Mey.), PI 319448 (V. adenantha G. Mey.), PI 93815 (Vigna angularis Willd.) and PI 157625 (V. angularis Willd.); the eight accessions of the second row are PI 360707 (V. angularis Willd.), PI 416742 (V. angularis Willd.), PI 527686 (V. angularis var. niponensis Owhi & H. Ohashi), PI 146800 (Vigna caracalla L.), PI 322588 (V. caracalla L.), PI 310294 (Vigna longifolia Verdc.), PI 292866 (Vigna luteola Jacq.) and PI 406329 (V. luteola Jacq.); the eight accessions of the third row are PI 164316 (Vigna mungo L. var. mungo), PI 208462 (V. mungo L. var. mungo), PI 218104 (V. mungo L. var. mungo), PI 305073 (V. mungo L. var. mungo), PI 292872 (Vigna oblongifolia A. Rich.), PI 181585 (V. oblongifolia A. Rich. var. oblongifolia), PI 292868 (V. oblongifolia A. Rich. var. parviflora) and PI 406358 (V. oblongifolia A. Rich. var. parviflora); the eight accessions of the fourth row are PI 164301 (Vigna radiata L. var. radiata), PI 171435 (V. radiata L. var. radiata), PI 381351 (V. radiata L. var. radiata), PI 427064 (V. radiata L. var. radiata), PI 240867 (Vigna subterranea L.), PI 245951 (V. subterranea L.), PI 378867 (V. subterranea L.) and PI 173933 (Vigna umbellata Thunb.); the eight accessions of the fifth row are PI 208460 (V. umbellata Thunb.), PI 247689 (V. umbellata Thunb.), PI 275636 (V. umbellata Thunb.), PI 291384 (Vigna unguiculata L. ssp. cylindrica), PI 292883 (V. unguiculata L. ssp. dekindtiana), PI 632910 (V. unguiculata L. ssp. pubescens), PI 215659 (V. unguiculata L. ssp. sesquipedalis) and PI 419163 (V. unguiculata L. ssp. sesquipedalis); the eight accessions of the sixth row are PI 610582 (V. unguiculata L. ssp. sesquipedalis), PI 632903 (V. unguiculata L. ssp. stenophylla), PI 632904 (V. unguiculata L. ssp. stenophylla), PI 582470 (V. unguiculata L. ssp. unguiculata), PI 582469 (V. unguiculata L. ssp. unguiculata), PI 582578 (V. unguiculata L. ssp. unguiculata), PI 612607 (V. unguiculata L. ssp. unguiculata) and PI 406390 (Vigna vexillata L.), respectively. The bar for seed size represents 1 inch.
There were five accessions investigated from V. angularis for phylogenetic analysis. One of them (PI 527686) was the wild form, which was different from the other four accessions (Fig. 2). The seed size of PI 527686 (2·37 g per 100 seeds) was almost four times smaller than the cultivated form (from 8·16 to 9·59 g per 100 seeds, Table 1). From phylogenetic analysis, three V. subterranea accessions clustered into two distinct groups. The accession (PI 378867) was distinct from the remaining accessions (PI 240867 and PI 245951) (Fig. 2). The seed morphology was consistent with the phylogenetic analysis. The former accession had a red seed coat, whereas the latter two accessions had reddish brown seed coats.
Phylogenetic analysis from DNA fragment data classified 12 investigated cowpea accessions into two small clusters. The first small cluster contained three accessions (PI 632904, PI 632910 and PI 632903) and the second small cluster contained nine accessions (from PI 291384 to PI 582578) (Fig. 2). All accessions within the first cluster had small seeds (1·23, 1·6 and 1·78 g, respectively) and this cluster may be called the ‘wild group’. All accessions except one within the second small cluster had large seeds (from 7·72 to 20·89 g) (Table 1 and Fig. 5) and this small cluster may be called the ‘domesticated group’. The result from seed morphological observation was consistent with those from phylogenetic analysis. Our results demonstrated that phylogenetic analysis with morphological re-examination may provide a more complete approach to classify accessions or to examine misidentified accessions in a plant germplasm collection.
Very limited genetic variation (especially diversity within a species) was detected within the USDA Vigna germplasm collection based on the present and previous studies. Currently, some genetic gaps exist in the USDA Vigna collection. For example, on the botanical variety level, there are at least three varieties (var. radiata, var. sublobata and var. setulosa) available within V. radiata (Bisht et al., Reference Bisht, Bhat, Lakanpaul, Latha, Jayan, Biswas and Singh2005). However, most accessions preserved in the USDA collection are from V. radiata var. radiata. Only one accession is maintained from V. radiata var. sublobata, while there are no accessions classified as V. radiata var. setulosa in the USDA collection. There are at least two botanical varieties (V. mungo var. mungo and V. mungo var. silvestris) within V. mungo (Seehalak et al., Reference Seehalak, Tomooka, Waranyuwat, Thipyapong, Laosuwan, Kaga and Vaughan2006), but only V. mungo var. mungo was collected and maintained as part of our collection. At the species level, some newly described ones (for example, V. aridicola, V. exilis, V. nepalensis, V. tenuicaulis and other species) are available (Tomooka et al., Reference Tomooka, Yoon, Doi, Kaga and Vaughan2002). Tomooka et al. found that the species V. aridicola was closely related to V. aconitifolia (moth bean), V. exilis was closely related to V. umbellata (rice bean), and both V. nepalensis and V. tenuicaulis were closely related to V. angularis (azuki bean). Although these newly described species could be potentially important for improving the cultivated species, none of them have yet been added to the USDA Vigna germplasm collection.
In conclusion, gene-derived markers are efficient to reveal phylogenetic relationships. Forty-seven Vigna accessions have been classified into four notable groups. Gene-derived markers are more effective at revealing polymorphism among species than within species. A few polymorphisms were identified within species by sequencing amplicons generated from gene-derived primers. The classification from DNA fragment analysis was consistent with the classification from DNA sequence analysis. Moreover, the genetic classification was supported by seed morphological observation. There was limited genetic diversity within the current USDA Vigna germplasm collection. In order to expand the genetic base of the USDA Vigna germplasm, new botanical varieties, subspecies and species need to be added to the USDA collection by germplasm curation and exchanges.
We are grateful to Drs Brad Morris, Richard Fery and Paul Raymer for comments and suggestions to improve the quality of the manuscript, Mrs Lee-Ann Chalkley for providing the seeds and Mr James Chalkley for help with scanning the seeds for morphological observation. We thank Mr Brandon Tonnis for his excellent technical assistance.
