A total of 28 accessions of Lima beans collected in the republic (Table 1) were analyzed. The study was carried out in the Laboratory of Molecular Biology and Bioinformatics Applied to Genomics at the National University of Sciences, Technologies Engineering and Mathematics of Abomey in Benin.

DNA was extracted from young leaves using the SDS method in accordance with [25] with minor modifications. Briefly, 500 mg of young leaves of Phaseoluslunatus were ground in liquid nitrogen in a mortar with 1000 µl of SDS buffer (200 mM Tris-HCl, 25 mM EDTA, 250 mM NaCl, 0.5% SDS). The ground material was incubated at 37˚C in a water bath for 1 hour. After cooling at room temperature, 800 µl of the phenol/chloroform/isoamyl alcohol (25:24:1) mixture was added to each sample and the whole was mixed before then being centrifuged at 15,000 rpm for 10 min at 4˚C. The supernatant was collected and mixed in equal volumes with a solution of Chloroform/isoamyl alcohol (24:1). The mixture was then centrifuged at 15,000 rpm for 10 min at 4˚C. The supernatant was once again collected in a new tube and the DNA precipitated with 0.1 volume

of sodium acetate (3M, pH 7.0) and 0.7 volume of cold isopropanol and left at −20˚C for 20 min. Final centrifugation of 15 min at 15,000 rpm 4˚C will be done to recover the pellet of ADN. The pellet of ADN was washed 3 times with ethanol at 70% and once with ethanol at 100% and dried at room temperature for about 2 hours and then suspended in 1X TE storage buffer.

DNA quality was checked on agarose gel at 0.8%. Its concentration was assessed with NanoDrop Lite (Thermo Fisher Scientific) and DNA purity was accessed using the absorbance ratio (A260/A280). Based on the obtained DNA concentration, different dilution rates were applied to each sample to obtain a concentration of 5 ng/μl necessary for a PCR amplification reaction.

A total of 10 microsatellite markers isolated and optimized for Common beans were used [26] (Table 2). PCR reactions were induced with 20 ng of DNA, 1 U of Taq DNA polymerase, 2.0 mM of magnesium chloride (MgCl₂), 0.2 mM of each dNTP, 0.1 μM of each primer, and 1X PCR reaction buffer in a final volume of 20 ml [26]. The program consisted of an initial denaturation at 94˚C for 2 min, followed by 45 cycles at 94˚C for 30 s, 49˚C for, 30 s and 72˚C for 30 s and a final extension at 72˚C for 10 min [11]. Amplification products were migrated on agarose gel at 2%, revealed with Ethidium Bromide, and then visualized on a UV transilluminator.

At each locus, different bands recorded an allelic composition. Thus, SSR alleles were coded as individual markers with 1 for the presence and 0 for the absence of the allele as binary data [27]. Genetic diversity parameters such as polymorphism rate (P), allelic diversity, and Polymorphism Information Content (PIC) were estimated. PIC value was calculated as followed: PIC = 1 − ∑ fi 2 : where fi is the frequency of each allele.

To assess genetic relationships between accessions, the dissimilarity between paired accessions was assessed using Darwin 6.0.21 software [28]. The generated dissimilarity matrix was used to infer a dendrogram using the Neighbor-Joining method.

The determination of the genetic groups was supported by population structure

analysis using the Bayesian model-based analyses using the software package Structure v2.3.4 [29]. The membership of each accession was run for the value of K = 2 to K = 10 with the admixture model and correlated allele frequency. Accessions with membership probabilities greater than 0.8 were grouped together and accessions with membership probabilities below 0.8 were assigned to the admixed group. For each K, it was replicated 3 times. Each run was implemented with a length of the burn-in period of 1000 followed by 10,000 Markov Chain Monte Carlo (MCMC) replicates. The ΔK method was used to identify the optimal value of K [30] using the online Structure Harvester program [31].

Principal Coordinate Analysis (PCoA) and Molecular Variance Analysis (AMOVA) were performed using GenAlEx 6.503 software [32] to estimate the genetic differentiation within and among the different subpopulations. Population genetic parameters such as the average number of total alleles (Na), Shannon-Weaver information index (I), the average number of private alleles (Pa), Shannon-Weaver diversity index (h), percentage of polymorphic loci, PhiPT and allele across the different subpopulations were calculated. Data were coded as suggested in the GenAlEx 6.503 software user manual [32].

SSR markers are nuclear markers that are widely used in population genetic studies and provide important information for the preservation of genetic diversity [33]. In the present study, of the 10 SSR markers, 6 were amplified in all accessions and were therefore retained for the various analyses. The 6 SSR markers analyzed generated a total of 28 alleles with an average of 4.16 alleles per SSR. The number of alleles ranged from 2 (BM160 and GATS91) to 8 (BM141). As repeat polymorphisms revealed by SSRs or RAPD and over DNA markers result from the addition or deletion of the entire repeat units or motifs [34], the polymorphism revealed by these types of markers depends both on the repeat pattern, its location in the genome of the species in connection with the speed of evolution of the species and on the size of the analyzed collection [35]. Thus, the relatively lower number of alleles reported in this study can be explained by the size of the analyzed collection, the location of the markers, or also, the number of markers used. The polymorphic information content value which indicated the informativeness of the loci and their ability to detect differences between the genotypes [36] ranged from 0.079 for the GATS91 marker to 0.680 for the BM 140 marker with an average of 0.408 (Table 3). This average PIC value is relatively low and supported the low level of discrimination of the SSR markers used in the present study however BM 140 marker was found to be the most appropriate for testing genetic diversity given its high level of polymorphism. These diversity indices which are the average number of alleles per locus and the PIC reported in the present study were slightly lower than the previously reported by Gomes et al. [26] who reported 10.27 alleles per locus and an average PIC value

of 0.675 studying a set of 153 lima bean accessions. These findings reinforced the fact that the size of the collection influences the number of total alleles.

However, these markers, although having presented a relatively low level of discrimination, made it possible to reveal using the Neighbor-Joining tree based on the genetic distance, 3 clusters within all the 28 analyzed accessions. Cluster 1 labeled in red grouped 8 accessions, cluster 2 labeled in black grouped 6 accessions, and cluster 3, labeled in blue grouped 14 accessions (Figure 1). This grouping of accessions revealed potential duplicates in the analyzed collection. Indeed, an analysis of the inferred tree showed 5 pairs of duplicates (V8V9, V3V20, V27V7, V5V23, and V26V24) which exhibited the same evolutionary distance. This may be due to usual genotype exchanges among farmers.

Revealing Lima bean population structure and genetic diversity is important for breeding efforts which is an essential step in any genetic improvement process through the identification of promising genotypes [37]. Data generated from SSR genotyping was used to assess the population structure of the Lima bean collection. The estimation of the delta K value, using Evanno’s method, showed the highest peak at K = 3 (Figure 2 and Figure 3), indicating that the 28 accessions could be grouped into three sub-populations based on differences in their genetic traits. The first one labeled in red grouped 9 accessions, the second labeled in green grouped 7 accessions, and the third labeled in blue groups 12 accessions. Our results contradicted Gomes et al.’s [26] study in the determination of the genetic diversity and population structure from a core collection of 153 Lima beans. Maybe to our SSR markers’ low level of discrimination. According to the probability of membership, the accessions of the different subpopulations were classified as pure or admixed. To this end, 25 accessions presented a certain belonging to one of the three sub-populations. Thus, the results of the structure revealed a low level of admixture of 10.71% suggesting a non-negligible differentiation between the different sub-populations [38].

Analysis of Molecular Variance results revealed significant differences within and between populations (p = 0.001). 32% of the genetic variation came from

inter-population and 68% of the genetic variation came from intra-population suggesting that the genetic variation within populations was larger than that between populations [39]. However, a high PhiP value (0.321) was found between the three sub-subpopulations indicating a high and significative (p = 0.001) genetic differentiation between these sub-subpopulations. Comparing genetic differentiation between sub-populations showed a low differentiation between sub-populations 1 and 3 (PhiPT = 0.25). The largest index was observed between subpopulation 2 and subpopulation 3 (PhiPT = 0.47). The high level of genetic differentiation found for the Lima bean in the present study may be explained by the reproductive characteristics of this crop. According to Penha et al. [11] and Nasir et al. [2], Lima bean had a mixed system with a predominance of self-fertilization, with only 38% of natural crossing which limits gene flow and therefore increases diversity among populations. This coincided with the AMOVA results (Table 4), where 32% of the total variation was accounted for by among-subpopulation variations. Our results are consistent with Martinez et al. [40] who reported low gene flow between populations of Lima beans from the Yucatan Peninsula which could be due to a low rate of seed exchange among farmers which could lead to low intermix among populations [2].

The allelic pattern and genetic diversity indices provided insight into genetic diversity within subpopulations [41]. The average number of alleles (Na) ranged from 3.83 within the accessions in sub-population 1 to 1.67 in sub-population 2. The percentage of polymorphic loci ranged from 50% to 100%. Subpopulation 1 had the highest average number of alleles (Na), Shannon-Weaver information (I), Shannon-Weaver diversity (h) indices (I = 1.18 and h = 0.64) and exhibited the highest mean number of private alleles (Pa = 0.11) (Table 4) meaning that subpopulation 1 was more diverse than the two overs. As a result, subpopulation 1 could be considered as the main gene pool and could serve as a source for the

selection of parents to improve the existing Lima bean landraces. The number of accessions with private alleles ranged from 4 in subpopulation 2 to 9 in subpopulation 1. The number of markers with private alleles ranged from 1 (for a total of 12 accessions) to 5 (for a total of two accessions: V27 and V7) (Table 5 and Table 6).

The first two axes of the PCoA expressed 43.83% of the total variation in the analyzed collection. The results revealed three well-separated clusters among the analyzed accessions in accordance with the population structure results (Figure 4) and coincided with the clustering based on the Neighbor-Joining tree.

SP: Subpopulation; N: Number of accessions; Na: Average number of alleles; I: Shannon-Weaver information index; h: Shannon-Weaver diversity index; Pa: Average number of private rare alleles; %P: Percentage of polymorphic loci.