Introduction

Cowpea, Vigna unguiculata (L.) Walp., is an important food legume in the tropics and subtropics, especially in sub-Saharan Africa, where it is grown for its seeds as well as for fodder (Ehlers and Hall 1997; Pasquet and Baudoin 2001; Timko et al. 2007; Xu et al. 2010; Boukar et al. 2016).

Cowpea gene pool organization is fairly complex, with numerous subspecies, including some of hybrid origin (Pasquet et al. 2021). This complexity may explain why genetic resources from the wild gene pool have never been used in breeding (Boukar et al. 2016). Indeed, although the different subspecies are morphologically well identified, there are numerous accessions that are intermediate or introgressed in various ways (Pasquet et al. 2021). Therefore, it would be helpful to identify molecular markers that can characterize subspecies, confirm morphological identifications or detect, qualify, and eventually quantify introgressions in some plants or accessions. Such molecular markers that could characterize the different subspecies are definitely needed to support the cowpea research community.

However, at a low taxonomic level, ancestral polymorphism or incomplete lineage sorting and hybridization do exist in plants, and information from all genomes is required (Naciri and Linder 2015). This is especially the case in cowpea. Chloroplast captures and lineage sorting were detected between cowpea infraspecific groups (Pasquet et al. 2021). Within cowpea, chloroplast DNA restriction fragment length polymorphisms (cpDNA-RFLPs) are interesting markers that can characterize the different subspecies (Pasquet et al. 2021), but this technique is obsolete and tedious, limiting its use. While the interesting restriction site mutations can be converted into much more convenient single nucleotide polymorphisms (SNPs), as Feleke et al. (2007) did for the BamHI s13.3 mutation, more cpDNA mutations should be found after full sequencing of the chloroplast genome of the various subspecies.

However, regarding the nuclear genome, there are no available markers for the molecular characterization of cowpea subspecies. Recent molecular research did not explore the cowpea gene pool beyond domesticated cowpea and its wild progenitor, i.e., subsp. unguiculata (Gupta et al. 2012; Huynh et al. 2013; Chen et al. 2017a; Otwe et al. 2017; Fatokun et al. 2018; Ketema et al. 2020; Gbedevi et al. 2021; Sarr et al. 2021; Munoz-Amatriain et al. 2021; Sodedji et al. 2021; Zuluaga et al. 2021; Dagnon et al. 2022; Gumede et al. 2022; Guimaraes et al. 2023). The only work focusing on the wild gene pool is that of Ogunkamni et al. (2008) based on simple sequence repeats (SSRs), but they did not try to characterize the different subspecies. However, SSRs were successfully used for the identification of closely related species in some complex taxonomic groups, e.g., Psidium (Tuler et al. 2015), Rhododendron (Wang et al. 2019), and Mediterranean Tamarix (Terrones et al. 2022).

Since several hundred cowpea SSRs are available, especially from functional regions (Timko et al. 2008; Andargie et al. 2014), the objective of the present work is to prove that SSRs can be used for molecular characterization of the nuclear genomes of cowpea subspecies. SSRs could improve the cowpea phylogeny, help cowpea gene bank managers characterize their wild cowpea accessions, and, ultimately, help breeders use wild cowpea genetic resources.

Materials and methods

Plant materials

The plant materials consisted of 30 wild cowpea accessions provided by Meise Botanical Garden, Belgium (http://db.plantentuinmeise.be/RESEARCH/COLLECTIONS/LIVING/PHASEOLUS/index.html), 3 wild accessions from Senegal and the breeding line Melakh provided by ISRA (Institut Sénégalais de Recherches Agricoles). All subspecies and taxonomic groups were represented, except subsp. dekindtiana (Harms) Verdc. sensu stricto from the South Angola Mountains, which is still absent from living collections (Table 1). Most of these accessions were already included in previous works (Pasquet 1999; Feleke et al. 2006; Pasquet et al. 2021), and the MT and SP numbers used previously were kept instead of their equivalent four-digit NI numbers from Meise Botanical Garden. Vigna vexillata (L.) A.Rich. NI 1014 was added as an outgroup. Plants were grown in pots filled with noninoculated sandy soil and watered with tap water twice a week.

Table 1 List of wild cowpea accessions
Full size table

DNA isolation and genotyping

The DNA extraction, PCR, and electrophoresis methods followed those of the Sarr et al. (2021) protocol. Considering the goal of our study, highly polymorphic SSRs were discarded, especially those showing polymorphism within subsp. unguiculata alone (Li et al. 2001; Diouf and Hilu 2005; Asare et al. 2010; Badiane et al. 2012; Ogunkamni et al. 2014; Ali et al. 2015; Desalegne et al. 2016; Xiong et al. 2016; Xu et al. 2010; Chen et al. 2017b; Desalegne et al. 2017; Sarr et al. 2021). A total of 61 SSR primers were selected and tested. The SSR primers can be downloaded from the Cowpea Genomics Knowledge Base (CGKB) (http://cowpeagenomics.med.virginia.edu/CGKB) (Timko et al. 2008).

Data analysis

Parsimony analysis was performed with Paup* 4.0a169 (Swofford 2017). The two most variable markers, i.e., SSRs 6193 and 6220, were removed from the dataset for this parsimony analysis.

Chromosomal location of the SSR markers and map construction

Each polymorphic SSR marker used in this study was blasted against the cowpea genome available in Phytozome (https://phytozome-next.jgi.doe.gov/). The markers were mapped to the chromosomes presented by Munoz-Amatriain et al. (2017) based on their physical position using MapChart 2.3 (Voorrips 2002).

Results

SSR polymorphism

Out of the 61 SSR primers tested, 27 yielded amplification products across all cowpea subspecies. Some primers, such as SSR 6326, amplified subsp. unguiculata and accessions from close subspecies but not the accessions of subspecies far from subsp. unguiculata, which suggests mutations in the anchoring region. The results from these primers were not included in the analysis.

Vigna vexillata was initially included as an outgroup, but the primers did not amplify the DNA for half of the accessions. For the other half, the V. vexillata allele was different from all the V. unguiculata alleles. The only exception was SSR 6209, which yielded an allele for NI 1014 that was similar to the allele of subsp. baoulensis. Therefore, NI 1014 was not included in the parsimony analysis, and the tree was not rooted.

Finally, 18 SSR markers were polymorphic (average 3.83 alleles per locus). With the exception of the very variable SSR 6193 (8 alleles) and SSR 6620 (12 alleles), the number of alleles varied from 2 to 5 for the polymorphic loci (Table 2). The 18 polymorphic SSRs were distributed among 10 chromosomes (Fig. 1). Some markers were located in close vicinity (SSR 6193 and 6222, SSR 6225 and 6246, SSR 6274 and 6674), but within these marker pairs, both markers behaved very differently.

Table 2 Genetic diversity information provided by the polymorphic markers used in this study
Full size table
Fig. 1
figure 1

Distribution of the 18 polymorphic simple sequence repeat (SSR) loci on 10 cowpea chromosomes

Full size image

Regarding SSRs that could be used for molecular characterization, i.e., that showed no variability within a subspecies or a group, 11 SSRs characterized 6 subspecies or varieties (Table 2). A combination of SSRs 6246, 6274, and 6920 characterized not only subsp. stenophylla but also SP 304. A combination of SSRs 6209, 6212, 6274, and 6920 characterized var. protracta. Var. protracta was the taxonomic group most difficult to characterize.

A unique combination of three alleles from SSRs 6246, 6274, and 7067 characterized most accessions from subsp. alba, subsp. tenuis, subsp. pubescens, and the BWA group of var. spontanea, as well as accession SP 141 from the IOCP group of var. spontanea.

Parsimony analysis

The parsimony analysis (Fig. 2) yielded numerous trees with a length of 52 single characters. They differed in the position of MT 340 (with subsp. pawekiae or with var. kgalagadiensis), SP 167 and SP 304, and SP 219 and SP 582 (with subsp. unguiculata, with the subsp. alba—subsp. pubescens polytomy, or in a fourth clade). The tree presented here has a consistency index of 0.6346 and a homoplasy index of 0.3654.

Fig. 2
figure 2

Parsimony analysis. SSR 6193 and SSR 6220 were not included in this parsimony analysis

Full size image

Although this tree is not rooted, we can consider a basal polytomy with 3 clades. The first clade includes subsp. baoulensis, subsp. letouzeyi, subsp. pawekiae, subsp. stenophylla, var. kgalagadiensis, and var. protracta, i.e., the main subspecies (Pasquet et al. 2021). The second clade includes subsp. pubescens, subsp. alba, the BWA group, and subsp. tenuis, i.e., the subspecies of hybrid origin (Pasquet et al. 2021). The third clade comprises subsp. unguiculata, including two accessions from the IOCP group.

Discussion

The SSRs tested are spread throughout the genome. They are not concentrated on a few chromosomes and are representative of the whole genome. The SSRs tested can characterize all the main subspecies (Pasquet et al. 2021), i.e., subsp. pawekiae, subsp. letouzeyi, subsp. baoulensis, var. protracta, var. kgalagadiensis, and subsp. stenophylla, as well as the annual subsp. unguiculata, but they failed to characterize most of the subspecies and groups of hybrid origin (Pasquet et al. 2021), i.e., subsp. alba, subsp. tenuis, and the BWA group and the IOCP group of var. spontanea. There is still no set of SSRs for characterizing subsp. tenuis or subsp. alba.

Although Pasquet et al. (Pasquet et al. 2021) performed parsimony analysis of cowpea chloroplasts, this is the first cowpea gene pool parsimony analysis based on nuclear DNA. The chloroplast DNA led to a seven-clade polytomy, while we observed a three-clade polytomy. Even though subsp. unguiculata formed a single clade in both analyses, there are major differences between the two analyses.

Chloroplast DNA clades A, B, D, and E and the accessions not belonging to any clade are here pooled into the main clade, with the exception of subsp. alba accessions, which here are included in the hybrid origin clade. Regarding the organization of the cowpea gene pool, this work confirms the opposition between the main subspecies and the subspecies of hybrid origin. With the exception of the paraphyletic subsp. stenophylla and var. protracta, all the main subspecies as well as the annual subsp. unguiculata are monophyletic. According to this nuclear phylogeny, var. kgalagadiensis could deserve subspecies status.

The split between the forest subspecies from the Mensensis group and the savanna subspecies from the Dekindtiana group does not appear in this analysis. The forest subspecies do not form a monophyletic group, nor do the savannah subspecies. Instead of the forest versus savannah opposition, there seems to be opposition between the main subspecies with a keel twisted toward the left (with the exception of subsp. letouzeyi) and the subspecies that show a keel twisted toward the right, i.e., subsp. unguiculata and the subspecies with a hybrid origin.

The subspecies of hybrid origin appear in a clade between the main subspecies and subsp. unguiculata, along with the BWA and IOCP groups. There are alleles (from SSR 6246, 6274 and 7067) which group all these accessions in this clade. Such a grouping did not appear in Pasquet (1999) or in Ogunkanmi et al. (2008). Although grouped by these SSR markers, these accessions belong to three different chloroplast clades (Pasquet et al. 2021). Chloroplast clades C and F are consistent with the present hybrid origin clade. Subsp. alba having a var. kgalagadiensis chloroplast but being located far from var. kgalagadiensis seems to be a clear example of old chloroplast capture. This confirms the hybrid origin of subsp. alba and suggests that the male ancestor capturing the var. kgalagadiensis chloroplast was subsp. tenuis (or a taxon close to subsp. tenuis) instead of subsp. unguiculata.

As observed with cpDNA (Pasquet et al. 2021), few accessions from the subspecies of hybrid origin were not in their expected clade. Subsp. tenuis MT 340 is associated with var. kgalagadiensis. It has 3 alleles in common with var. kgalagadiensis and 3 alleles in common with the other subsp. tenuis accessions. Subsp. tenuis SP 304 is also misplaced due to its allele at SSR 6246, which is mainly observed in var. protracta (SP 304 was collected in Port Saint Johns in South Africa, a few kilometers away from a var. protracta area). Similarly, SP 141 is close to subsp. alba, subsp. tenuis and subsp. pubescens due to its allele at SSR 6246. These accessions are from a geographic area where different subspecies are known to overlap and where numerous intermediate plants are encountered. These discrepancies are likely due to recent hybridizations or to incomplete lineage sorting (Naciri and Linder 2015).

This work also confirms the special status of the annual subsp. unguiculata. In all the analyses, subsp. unguiculata was separated from the different perennial subspecies. This can be explained by its annual status. More generations should lead to the accumulation of more mutations, as observed previously with cpDNA (Pasquet et al. 2021). This should contribute to the isolation of this subspecies in the different analyses.

Var. protracta, located at the bottom of the clade including all the main subspecies in the parsimony analysis and not as well grouped as the other main subspecies, appears to be the oldest subspecies. Since the parsimony analysis tree is not rooted, we could also consider var. protracta as a hinge between the main subspecies and the group composed of subsp. unguiculata and the subspecies of hybrid origin. This should be in agreement with the hypothesis that the species Vigna unguiculata originated in southern Africa (Padulosi 1993).

Conclusion

Unfortunately, subsp. dekindtiana sensu stricto from southern Angola is still unavailable, and the outgroup accession was too distantly related, which hampered the reconstruction of the complex evolutionary history of V. unguiculata. However, this work can be considered the first attempt to perform parsimony analysis of the V. unguiculata nuclear genome.

Of course, a larger set of primers would need to be tested on a larger set of accessions, but the SSRs tested allowed us to characterize subsp. pubescens and all the main subspecies (Pasquet et al. 2021).

We can conclude that SSR markers from functional regions are an ideal tool for cowpea subspecies molecular characterization, especially since SSR analyses can be multiplexed (e.g., Mitchell et al. 1997). SSRs are robust and very reliable molecular markers that are widely used in cowpea, and they are most cost effective than sequencing. They do not require costly equipment or bioinformatic skills. In addition, since subsp. dekindtiana sensu stricto accessions are still unavailable, it is too early to develop several thousand SNPs based on next-generation sequencing. In the meantime, as long as subsp. dekindtiana sensu stricto accessions continue to be unavailable, SSRs could be used on a large scale for characterizing wild cowpea accessions in different gene banks.

Associated with SNPs derived from chloroplast restriction site mutations, such a tool should help understand the complex evolutionary history of the cowpea gene pool as well as improve its taxonomy. Perhaps more importantly, it should help breeders access the greatest part of the cowpea gene pool diversity.