Introduction

The lima bean (Phaseolus lunatus L.) is an important species of the Fabaceae family distributed worldwide (Baudet 1977). It provides an alternative source of proteins, fibers, and mineral elements for the growing population (Seidu et al. 2018). Studies conducted by Chacón-Sánchez and Martínez-Castillo (2017) reported that wild lima bean is classified into three major gene pools according to genomic data: two Mesoamerican (MI and MII) and one Andean (AI) gene pool. The possible existence of another Andean gene pool, the AII gene pool restricted to the Andes in central Colombia, has been proposed, although this has not been confirmed.

Genetic variability in the species largely relates to the process of global dispersion, which requires adaptation to different environments, making it a promising crop for improving food security under the predicted scenarios of climate change in Latin America and elsewhere (García et al. 2021). Conservation of the genetic resources of the lima bean is conducted ex situ in germplasm banks.

According to Govindaraj et al. (2015), despite the high availability of genetic diversity, managing and exploring a large number of accessions in collections can be challenging. The size of the collections, presence of duplicates, limitation of financial resources, and information about the traits of each conserved accession have contributed to a disparity between the growth of these collections and their effective use (Brown 1989a; Xu et al. 2016). To overcome this, the establishment of core collections (CCs) has been proposed, which can be defined as a small set of accessions representative of the genetic variability of a germplasm collection, with minimum redundancy and appropriate for curators, breeders, and geneticists (Frankel and Brown 1984). According to Brown and Spillane (1999), the steps for selecting a CC are (1) identifying the material that will be represented, (2) determining the size of the core collection, (3) dividing the accessions used into different groups, (4) determining the number of entries per group, and (5) choosing the entries from each group that will be included in the CC. Several descriptors can be used in combination for the implementation of a CC, and the most common are passport data; geographic origin; morphological, phenotypic, biochemical, and phyto-pathological characteristics; and molecular markers (Balfourier et al. 2007; Bhattacharjee et al. 2007; Kumar et al. 2016; Martins et al. 2015).

Studies have shown the importance of seed traits in defining gene pools and estimating the genetic diversity of lima bean crops (Chacón-Sánchez and Martínez-Castillo 2017; Silva et al. 2017, 2019). Gomes et al. (2020), using 153 Brazilian accessions, proposed the use of microsatellite data for the establishment of a lima bean CC for the germplasm bank of Phaseolus at the Universidade Federal do Piauí (GBP-UFPI). However, through exchanges, the GBP-UFPI recently acquired accessions of lima beans with different origins from several countries (Andean, Mesoamerican, and South American, among others), including wild genotypes. With the incorporation of new accessions, the question arises of the need to update the core collection of lima beans in Brazil to maximize genetic diversity and ensure a balanced representation of accessions from various collection sites, with minimal redundancy. Evaluating genetic diversity differences between the core collections (CCs) and the entire collection (EC) is a useful approach for testing the effectiveness of the core collection. Various measures, such as genetic diversity parameters, have been used to evaluate and validate the CC (Meryem et al. 2023). Based on the available literature, this is the first study on a CC of this species that is compiled of the integration of geographic distribution, phenotypic, and molecular data. In this context, the main objective of this study was to construct the CC of lima bean in Brazil from the GBP-UFPI, based on two selection strategies that integrate different types of data, and to perform validation based on nuclear microsatellites (ncSSRs) to optimize this germplasm conservation.

Materials and Methods

Germplasm

This study was performed with 395 lima bean accessions from the GBP-UFPI, located in Teresina, Piauí, Brazil. All accessions were characterized based on 11 phenotypic seed descriptors in the Genetic Resources and Plant Breeding Laboratory, Plant Science Department, UFPI.

Eight quantitative and three qualitative seed traits were evaluated in the entire collection (EC) and were described using lima bean crop descriptors from the International Plant Genetic Resources Institute (IPGRI 2001). The quantitative descriptors were (a) seed area (SA, mm2), (b) seed length perimeter (SLP, mm), (c) seed length (SL, mm), (d) seed width (SW, mm), (e) seed length to width ratio (SL SW−1), (f) seed weight (100 WS, g), (g) seed thickness (ST, mm), and (h) seed thickness-to-width ratio (ST SW−1). The Smartgrain software (Tanabata et al. 2012) was used to determine the measures. Seed thickness was measured using a digital caliper. The multi-categorical qualitative descriptors evaluated were (a) seed background color (SBC), (b) standard seed color (SSC), and (c) second standard seed color (SSSC).

Procedures to Establish the Core Collection of Lima Bean

The establishment of the lima bean CC was divided into four stages (Fig. 1): (1) choice of the sample size, in which 79 accessions was considered an adequate size to be managed by the curators at a low relative cost, representing the sampling intensity of 20%; (2) use of the stratified sampling strategy, whereby the lima bean accessions were divided into four groups according to the centers of domestication and diversity: Brazilian (295 accessions), Mesoamerican (48 accessions imported from Costa Rica, El Salvador, Guatemala, Honduras, and Mexico), Andean (12 accessions imported from Colombia, Ecuador, and Peru), and Other (40 accessions imported from Azerbaijan, China, the Philippines, Ghana, Nigeria, and the USA); (3) determination of the proportion with which each group was represented, using either the proportional strategy (P), where of each group was represented in the CC in proportion to its size in the entire collection or the logarithmic strategy (L), in which the number of accessions to sample each group was proportional to the logarithm of the group size (Vasconcelos et al. 2010); and (4) selection of the accessions of each group to be included in the CC using the inverted Tocher’s method of grouping accessions with greater dissimilarity (Nascimento et al. 2018; Vasconcelos et al. 2007).

Fig. 1
figure 1

Scheme of procedures to establish the core collection of lima bean

Full size image

Evaluation of Candidate Core Collections with the Entire Collection

The validation process for the proportional (P) and logarithmic (L) core collections involved comparing them with the entire collection (EC) from which they were developed. Comparisons between were made of the quantitative traits using statistical parameters of amplitude, mean, variance, kurtosis, skewness, and average variability retention index (VRI), which was calculated according to Diwan et al. (1995). Comparisons of variances and means between the EC and core collections were performed using the F-test and Student’s t-test, respectively (Snedecor and Cochran 1980). The Shannon index was used to assess multi-categorical qualitative traits (Shannon and Weaver 1949).

Comparison of Core Collections Based on Molecular Data

A comparison was made of the disjoint set between the P and L (DS) CCs using nuclear microsatellites (ncSSRs). DNA was extracted from young lima bean leaflets using the modified CTAB method (Doyle and Doyle 1987) at the Genetic Diversity and Breeding Laboratory of the University of São Paulo, “Luiz de Queiroz” College of Agriculture (USP/ESALQ). DNA quality and quantity were checked by electrophoresis in 1% agarose gel with SYBR safe and compared with standard phage lambda DNA (Gomes et al. 2020).

A set of 13 ncSSRs were used: BM140, BM141, BM143, BM146, BM156, BM160, BM164, BM170, BM183, BM189, BM211, BM212 (Gaitán-Solís et al. 2002), and FF01 (Penha 2018). Simple sequence repeat (SSR) primers were synthesized using an M13 tailed in the forward primer. The ncSSRs were amplified as described by Penha (2018), and genotyping was performed using an automated DNA sequencer ABI 3500xL (Applied Biosystems). The sizes of the SSR fragments were analyzed and determined using the GeneMarker program (Applied Biosystems).

Genetic diversity was described in terms of the average number of alleles per locus (Na), percentage of polymorphic loci (%P), number of private alleles to the group (Pa), coefficient of inbreeding (F), observed heterozygosity (Ho), expected heterozygosity (He), results of principal coordinate analysis (PCoA), and analysis of molecular variance (AMOVA), using the GenAlex v. 6.5 software (Peakall and Smouse 2012).

The relationship between clusters among accessions in the DS was investigated using a dendrogram based on Rogers’ distances (1972). The dendrogram was built using the unweighted pair group method with arithmetic means (UPGMA), and the confidence of relationships was assessed with a bootstrap procedure with 1000 resamplings. MEGA software (v. 6.0) was used to edit the final tree (Tamura et al. 2013).

To assess the genetic structure existing among accessions in the DS, a Bayesian clustering algorithm was used with the program STRUCTURE 2.3.3 (Pritchard et al. 2000). The program was executed with the number of clusters varying from one to ten. For each run, 200,000 burn-in and 500,000 Monte Carlo Markov chain (MCMC) iterations were performed. The admixture model with correlated allele frequencies was used. The most likely K number of clusters was determined by Evanno et al. (2005) based on delta K, calculated using the application Structure Harvester (Earl and Vonholdt 2012).

Results and Discussion

Validation of the Representativeness of Strategies to Form the Core Collection

From the EC of 395 lima bean accessions, two CCs were established of 79 accessions each, through stratified sampling (Table 1). The number of accessions in a core collection is an important determining factor of its effective use (Brown and Spillane 1999). The size reduction allows for efforts to be concentrated on the characterization and evaluation of germplasm to form a more complete information base for this set of accessions. The size reduction also cuts costs and allows resources to be directed to other activities, such as germination tests, regeneration, and molecular analysis, facilitating user access to the germplasm collection. The previously developed lima bean CC consisted of 34 accessions (Gomes et al. 2020), whereas the CCs developed in the present study were comparatively higher because of the larger number of accessions in the initial germplasm collection.

Table 1 Number of accessions selected in each group using the inverted Tocher’s method
Full size table

The proportion with which each group was represented was based on the P and L sampling strategies, in which 49 accessions coincided. The 30 accessions that differentiated the proportional CC were all from the Brazil group, which had the highest number of accessions. However, the L strategy allowed for the groups with fewer accessions, the Mesoamerican, Andean, and Other groups, to be more represented in the CC. Thus, the L strategy decreased the weight given to large groups compared to that of the P strategy (Oliveira et al. 2010).

Once established, each CC was investigated for its ability to retain the EC variability. Table 2 summarizes the comparison between EC and CCs. When comparing the EC with the CCs regarding the values of amplitudes and means, no significant differences were observed for the means of the traits, except for seed thickness in the logarithmic CC. The P and L strategies maintained similar values to the EC but were unequal in variance (variance of the EC < variance of the CC) and kurtosis (kurtosis of the EC > kurtosis of the CC), which indicates differences between the distributions. According to Van Hintum et al. (2000), maintaining genetic representativeness in the core collection implies the maintenance of the mean and amplitude, and expansion of phenotypic variance among individuals in the core collection in relation to the EC, which occurred in the two established CCs. The same pattern was observed in comparisons with the average variability retention index, whereby strategy P (99.32%) did not differ from strategy L (99.62%). Sampling is efficient when the average variability retention index is greater than 80% (Frankel and Brown 1984). Regarding the Shannon index, it was observed that for the qualitative traits related to seed color, the established CCs (P and L) showed values similar to or close to those of the entire collection (EC), which indicated that a good portion of the variability related to seed colors was maintained (Table 3). In summary, these two strategies were efficient in retaining variability when considering phenotypic traits.

Table 2 Minimum amplitude (Min) and maximum amplitude (Max), mean, variance (Var), kurtosis (Kur), and skewness (Sk) of the eight quantitative traits evaluated in the entire collection (EC) of the Active Germplasm Bank of Phaseolus, from the Federal University of Piauí (GBP-UFPI), and the two proportional (P) and logarithmic (L) core collections
Full size table
Table 3 Shannon indexes for qualitative traitsa of the entire collection and two core collections
Full size table

Comparison Between Core Collections Based on Molecular Data

The use of molecular markers, which directly reflect variation at the genetic level, is appropriate because each accession chosen to make up the CC is based on its allelic potential, regardless of its phenotypic expression. However, methodologies based on molecular markers are expensive, labor-intensive, and require knowledge and equipment that are not always available (Mahmoodi et al. 2019). Therefore, the use of molecular markers only for the validation of CCs established from large germplasm collections through phenotypic traits reduces costs and ensures the selection of the best CC strategy.

The 13 SSR markers analyzed in the comparison of the accessions that did not coincide in the P and L CCs generated 58 alleles, with an average of 4.46 alleles per locus, ranging from two (BM146, BM164, BM189, and FF01) to nine alleles (BM140) (Table 4). The mean values of observed heterozygosity (Ho) were generally lower than those of expected heterozygosity (He), except for the BM160 locus, resulting in high inbreeding coefficients (F). Martínez-Castillo et al. (2006) also found an excess of heterozygotes in the BM160 locus in wild lima bean populations in the Yucatan Peninsula, Mexico. The results of low Ho values were expected since lima bean is a mixed species with a predominance of autogamy (Penha et al. 2017). According to Burle et al. (2010), in autogamous species such as Phaseolus beans, low Ho values are expected because most of the loci are homozygotes. In lima beans, low Ho values were also observed by Chacón-Sánchez and Martínez-Castillo (2017), when comparing domesticated accessions with their wild ancestors, with means of Ho = 0.00 and He = 0.25. Silva et al. (2019) found low Ho (ranging from 0 to 0.065, with a mean of 0.025) in 183 lima bean landraces collected from different regions of Brazil. Andueza-Noh et al. (2015) also found low Ho (0.040), but high He (0.667), along with a high F (0.943), in 155 domesticated and wild lima bean accessions.

Table 4 Estimates of genetic diversity for the accessions that did not coincide in the proportional (P) and logarithmic (L) CCs, based on 13 SSR markers. Mean number of alleles (Na), observed heterozygosity (Ho), expected heterozygosity (He), and inbreeding coefficient (F)
Full size table

Comparisons of the estimates of genetic diversity (%P, Na, and He) revealed that the values for the non-coincident accessions of the logarithmic CC were higher than those of the proportional CC (Table 5). These results indicate greater genetic diversity in the logarithmic CC. The explanation lies in the fact that the non-coincident accessions of proportional CC were all from the Brazil group; hence, they are Brazilian and domesticated accessions. Conversely, the non-coincident accessions of the logarithmic CC were part of the Mesoamerican, Andean, and Other origin groups and had a domesticated and wild biological state. The high value of F (0.808) found in strategy P (Table 5) is justified by the founding effect and the fact that they were all domesticated accessions, which was not the case with strategy L.

Table 5 Estimates of genetic diversity for the accessions that did not coincide in the proportional (P) and logarithmic (L) CCs, based on 13 SSR markers, considering the percentage of polymorphic loci (%P), average number of alleles per locus (Na), number of private alleles to the group (Pa), observed heterozygosity (Ho), expected heterozygosity (He), and inbreeding coefficient (F)
Full size table

The logarithmic CC had a higher number of private alleles than the proportional CC (Table 5), which demonstrates that strategy L reduces the weight given to groups with a large number of accessions that may have a higher level of genetic redundancy. This information can be used to increase the probability of sampling rare alleles that confer broad or localized adaptation. According to Brown (1989b), strategy L is superior to strategy P if the degree of genetic variation between and within a group is unknown. If stratification is successful in concentrating the alleles of interest, then the logarithmic scale is a good way to recover alleles present in different distribution patterns. In the case of unsuccessful stratification, the two strategies (P and L) are equally appropriate and slightly less efficient than simple random sampling.

The greater genetic diversity of the logarithmic CC was also represented by the PCoA since it showed a greater distribution of the accessions in the first two coordinates (Fig. 2), which accounted for 41.76% of the variation and allowed for the observation of an overlap of some accessions belonging to logarithmic CC with the accessions selected for proportional CC. This result suggests that the logarithmic CC contains much of the genetic diversity contained in the proportional CC. Analysis of the accessions in the proportional and logarithmic CCs using the UPGMA method (Fig. 3) showed clusters similar to the PCoA. The accessions of the logarithmic CC were distributed throughout the dendrogram; that is, there were representatives of the accessions of the logarithmic CC in almost all groups formed by the analysis. In contrast, the proportional CC accessions were divided into two distinct clusters.

Fig. 2
figure 2

PCoA plot of accessions that do not coincide with the proportional CC (blue rhombuses) and logarithmic CC (orange squares) based on the variation of 13 microsatellite loci

Full size image
Fig. 3
figure 3

UPGMA dendrogram constructed with the distances of Rogers (1972), illustrating the clustering of accessions that did not coincide in the proportional CC (blue rhombuses) and logarithmic CC (orange squares), based on the variation of 13 microsatellite loci

Full size image

AMOVA showed that only 8% of the variation occurred between the CCs, indicating that most of the genetic variability was maintained in the CCs. Thus, most of the variations were within each CC (Table 6).

Table 6 Analysis of molecular variance (AMOVA) considering the accessions of the proportional (P) and logarithmic (L) CCs
Full size table

The likely number of genetic clusters was K = 2, suggesting the use of the ΔK method (Fig. 4), showing the formation of two main groups. However, a small peak was also observed at K = 5. The representation of the bar graph for K = 2 and K = 5 of the Bayesian analysis of STRUCTURE (Fig. 5) showed a complex ancestry relationship, which may be partly caused by the gene flow that occurs freely between wild and domesticated accessions, evidenced mainly in the accessions of the logarithmic CC since the proportional CC contained only Brazilian and domesticated accessions. Considering K = 2 (Fig. 5) in the proportional CC, the formation of two distinct genetic groups was observed. Gomes et al. (2020), when analyzing the structure of genetic variability through Bayesian analysis among 153 Brazilian bean accessions, also found different genetic groups. Logarithmic CC can be considered the preferred collection for the conservation and characterization of lima beans since it showed greater genetic variability than proportional CC.

Fig. 4
figure 4

Graphic representation of the optimal K value using 13 SSR loci of the accessions that did not coincide in the proportional (P) and logarithmic (L) CCs obtained by the program STRUCTURE inferred using the criterion ΔK of Evanno et al. (2005)

Full size image
Fig. 5
figure 5

Bar graph in K = 2 and K = 5 of the Bayesian analysis obtained by the program STRUCTURE, showing association coefficients for accessions that did not coincide in the proportional (P) and logarithmic (L) CCs. Accessions are ordered according to the CC to which they belong. Different colors represent distinct genetic groups

Full size image

Overall, we report that the established lima bean core collection provided by this study will be essential for improving conservation, the use of genetic resources, and the efficiency of the breeding program for the species. The demand for information and seeds from the accessions that make up the CC will likely increase. Given this perspective, procedures for regeneration and multiplication must be adopted. Moreover, it is worth highlighting that the constitution of the CC should not be considered permanent; instead, it should be dynamic, evolving with the evaluation of new accessions and the generation of new information. Future studies should involve the establishment of thematic core collections, such as those focusing on drought tolerance and pathogen resistance, in the face of climate change.

Conclusion

The present study is the first to use molecular diversity data in conjunction with phenotypic data and geographic distribution to establish a CC for lima beans. The small size of the CC is advantageous for field studies and mapping of genomic associations. Logarithmic CC can provide access to genetically diverse and agronomically important germplasms, which can be useful for broadening the genetic basis of the crop. The use of data of different types should be prioritized when available to establish the core collection. The morpho-agronomic data had low efficiency in establishing differences between the two CCs but were fundamental for the initial selection of accessions. Logarithmic CC proved to be more suitable for conservation purposes and should be examined in detail in the establishment of thematic core collections for the species.