Common bean is the most widely consumed legume species of the genus Phaseolus (Freytag and Debouck, 2002).  It is a pulse crop used since pre-Columbian times in the Americas and, since the 16^th century, in other regions of the world (Gepts et al., 2008). It is a true diploid (2n = 2x = 22) with a small genome (580 Mbp; Broughton et al., 2003). Originating in the Neotropics, common bean was domesticated in Mesoamerica and the Andes (Gepts and Bliss, 1986; Gepts, 1988). The crop has high diversity that is broadly classified into six or seven domesticated races distributed into two gene pools (Singh et al., 1991a, b, c; Blair et al., 2007, 2010b; Pallottini et al., 2004; Kwak and Gepts, 2009; Kwak et al., 2012). The crop is a major legume in Eastern and Southern Africa, occupying more than 4 million ha annually and providing food for ≥100 million people (Wortmann et al., 1998; Fisseha, 2015).  Of the total production in sub-Saharan Africa of over 3.5 million MT, 62% is in Eastern and Central African countries (Wortmann et al., 1998; Fisseha, 2015). Common bean became established with the African-European trade, even before the widespread era of colonization (Allen and Edje, 1990; Asfaw et al., 2009). Historical accounts show that common bean was introduced to Ethiopia in the 16^th century by Portuguese traders and rapidly became an important component of the diet there (Assefa, 1985; Fisseha, 2015). Ever since the introduction of common bean into Ethiopia, farmers have developed farming practices adapted to local conditions by preservation and exploitation of useful alleles, which have resulted in a range of morphologically diverse landraces (Sperling, 2001). Moreover, recent efforts of the national bean-breeding program in Ethiopia have targeted improvement of on-farm common bean productivity and have benefited since the 1980’s from continuous introduction of new germplasm from different parts of the world (Fisseha, 2015).

Today, Ethiopia is among the major bean producers in Sub-Saharan Africa (Wortmann et al., 1998). However, the national bean yield still lags behind the global average (Fisseha, 2015). This can be attributed partially to the low yielding capacity of cultivars under use (Assefa, 1990; Fisseha, 2015). To this end, it is essential to tap the potential of landrace genetic resources in order to introgress novel genes of adaptation, resistance to diseases and pests, and tolerance to abiotic stresses. According to Hornakova et al. (2003), landraces grown by small farmers are rich sources of valuable genes.

East Africa is a secondary center of diversity for common beans, due to the wide range of landraces there (Martin and Adams, 1987; Asfaw et al., 2009, Blair et  al.,2010b). Understanding the patterns and levels of genetic diversity of bean landraces and cultivars can shed light on potential adaptation and direction and level of gene flow, and eventually help bean breeding and conservation in Ethiopia. Hence, this research project was undertaken with the principal goal of evaluating the genetic diversity within and between common bean landraces, to classify genotypes based on clustering and to understand the distribution of genetic diversity between and within ecological/geographical regions of Ethiopia.

Plant materials

A total of 116 landrace accessions collected from a range of common bean production agro-ecologies in Ethiopia, four Ethiopian cultivars, three Kenyan cultivars, and two other cultivars, used as control genotypes for the Andean and Mesoamerican gene pools, respectively, were grown in August, 2012, in a greenhouse in the Biosciences Eastern and Central Africa (BecA-ILRI) hub in Nairobi, Kenya, for DNA extraction and analysis. The Ethiopian accessions were sampled from potential bean growing areas in the country (Supplementary Table 1 and Figure 1). The seeds of the control and commercial cultivars were acquired from the Ethiopian National Bean Research Project, based at Melkassa Agricultural Research Center, Adama, Ethiopia. The landrace accessions were provided by the Gene Bank of the Ethiopian Biodiversity Institute (EIB). A total of ten plants per each accession were planted in a single row in the screen house of BecA-ILRI hub, Nairobi, Kenya in August, 2012 for DNA extraction.

Genomic DNA extraction

For the molecular diversity assessment, total genomic DNA for each accession was isolated from a bulked leaf tissue sample of one week old plants from five randomly selected plants per accession using cetyltrimethylammonium bromide (CTAB) method (Doyle and Doyle, 1987) with some minor modifications, as described in supplementary part 1. However, 47 accessions did not produce enough genomic DNA, probably due to poor leaf sample qualities, which, in turn, imposed the need to  repeat DNA extraction from the same, using the Zymoplant seed DNA extraction kit (descriptions on the protocol are presented in Supplementary Part 2).

Microsatellite amplification

Twenty-four (24) microsatellite markers from all the 11 linkage groups were selected based on their Polymorphic Information Content (PIC) values and dispersed map locations (Yu et al., 2000; Pedrosa-Harand et al., 2008; Kwak and Gepts, 2009). Out of the 24 SSR markers, 15 were genomic, and the remaining nine were non-genomic (genic) markers (Supplementary Table 2). Markers were PCR amplified with 6-FAM, NED, PET or VIC 5’-labeled forward primers and unlabeled reverse primers. The primers were run in multiplexes, based on their fluorescence dye and allele size using BIONEER ACCUPOWER® Multiplex PCR Premix Kits (Supplementary part 3). Out of the 24 SSR markers, seven were dropped after preliminary evaluation, because they either produced no amplification (BM172 and BMd1) or were monomorphic (BM188, BM183, BMd16, PV-AG001, and PV-AT001). PCR products were run on an ABI PRISM 3730xl fragment analyzer (Applied Biosystems, Foster City, CA, USA) at the BecA-ILRI hub (Sequencing, genotyping, and Oligo unit, SegoLip), and allele sizes were determined by comparing with Genescan LIZ500 size standard using GeneMapper v. 3.7.3.7 software. The observed allele sizes were then adjusted for the discrete allele size using the AlleloBin software (http://test1.icrisat.org/gt-bt/download_allelobin.htm).

SSR genetic diversity analysis

Genalex 6.5b3 (Peakall and Smouse, 2012; http://biology.anu.edu.au/GenAlEx/) was used to calculate genetic diversity parameters, such as genetic distance, number of alleles (N_a); number of effective alleles (N_e); number of private alleles (N_pa); observed heterozygosity (H_o); expected heterozygosity (H_e); Shannon’s information index (I); analysis of molecular variance (AMoVA); and principal coordinate analysis (PCoA). Genetic associations were determined using the neighbor-joining coefficient with Darwin V. 5.0 (http://darwin.cirad.fr/darwin). Genepop V.4 (Rousset, 2008) and Popgene32 (Yeh et al., 1999) programs were also used to determine genetic diversity, polymorphic loci, gene flow, levels of heterozygosity, fixation index, and F-values,. Finally, PowerMarker v. 3.25 (Liu and Muse, 2005) was used to estimate the number of alleles, polymorphic information content (PIC) values, genetic distance matrices, observed heterozygosity (H_o); and  expected heterozygosity (H_e) for each marker across all genotypes and then across genotypes within and between gene pools.

Analysis of population structure

The software program STRUCTURE was run for K values ranging from 2 to 8. Each run was performed using the admixture model and 5,000 replicates for burn-in and 50,000 during the analysis (Pritchard et al., 2000). Evanno et al. (2005) test was performed after 10 simulations per K value. The repeated simulations were conducted for every subpopulation number from K = 2 to K=8 using 5,000 replicates for burn-in and 50,000 replicates according to previous suggestions (Rosenberg et al., 2002; Evanno et al., 2005; Ehrich, 2006).  The Δ statistic showed that K = 5 was the optimal number of subpopulations in this analysis (Supplementary Figure 1). This ideal K value presented the highest peak for change in value from and to the previous and subsequent numbers of subpopulations, respectively. This showed a gain in precision from subdividing the genotypes into five subpopulations versus any lower or higher numbers of subpopulations. The K=2 analysis was done with a particular interest of distinguishing between Andean and Mesoamerican accessions (Koenig and Gepts, 1989; Kwak and Gepts, 2009). To this end, five independent runs were performed with the admixture model and 5,000 replicates for burn-in and 50,000 replicates during analysis. The clustering in different runs was almost identical (similarity coefficient 0.9914). The run with the lowest likelihood value was selected among the five runs, and the accessions with more than 80% posterior assignment probability in the Mesoamerican cluster were assigned to the Mesoamerican gene pool (and vice versa for the Andean gene pool) (Supplementary Table 3). Lower posterior assignment probability values (that is, between 50 and 80%) may actually indicate hybrids or admixed accessions rather than ‘‘pure’’ accessions (Kwak and Gepts, 2009). Nonetheless, such accessions were included in the K=2 analysis, as they are important in future studies towards shedding light on the population structure of the common bean in Ethiopia, and as baseline information in breeding/improvement programs.

The hierarchical classification scheme into subpopulations comprised of Andean and Mesoamerican genotypes obtained here was in agreement with that reported for common bean germplasm in various studies (Singh et al., 1991a; Gepts, 1998; Diaz and Blair, 2006; Blair et al., 2007, 2010a, 2011; Okii et al., 2014b). Moreover, the moderate to mostly large differentiation among subpopulations (F_st values) were higher than was reported in other studies (Asfaw et al., 2009; Okii et al., 2014b). On the other hand, the higher differentiation recorded in the present study among Mesoamerican subpopulations compared to their Andean counterparts was in contrast with the results of Asfaw et al. (2009) and Okii et al. (2014b). The separation of bean accessions into the two gene pools was also evidenced in the NJ and PCoA analyses. Five cluster groups were identified, which supports the findings reported in previous studies (Kwak and Gepts, 2009; Burle et al., 2011). Furthermore, the presence of moderate admixture level agrees with previous reports (Asfaw et al., 2009; Blair et al., 2010b; Okii   et   al.,   2014b).   Similarly,    the   concurrence    ofSTRUCTURE results with that of PCoA and NJ analyses was in agreement with that reported by Asfaw et al. (2009) and Okii et al. (2014b). 

Five subpopulations with moderate admixture level and some switching of membership were observed in the present study. In line with this, Okii et al. (2014b) noted that the high level presence of admixture is indicative of the considerable mixing of common bean germplasm in planting and consumption and in hybridization in breeding. The considerable presence of admixture and switching of membership in some instances was supported by the PCoA and NJ analyses. These agree with some previous reports (Asfaw et al., 2009; Blair et al., 2010b; Burle et al., 2011; Okii et al., 2014b). In addition, the PCoA and NJ analyses in terms of geographic/ecological sampling of accessions indicated that accessions from different collection sites clustered together, which implied there was significant exchange of planting materials among farmers in different growing regions in the country. Moreover, the analysis of hybrid/ non-hybrid accessions indicated the Mesoamerican genotypes had higher instances of hybridity than the Andean counterparts. This observation supported previous results (Asfaw et al., 2009; Kwak and Gepts, 2009). On the other hand, from a population differentiation viewpoint, Andean genotypes were more differentiated than those from the Mesoamerican gene pool, which concurs with the findings reported previously for East African common bean germplasm (Asfaw et al., 2009; Okii et al., 2014b). 

Accessions with Andean origin had higher allelic parameter values (Na, Ne, PIC, etc.) than Mesoamerican accessions. This contrasted with results from other related studies (Asfaw et al., 2009; Burle et al., 2011; Okii et al., 2014 a,b). Such differences may be attributed to differences in genetic samples and respective sampling methods employed. On the other hand, the hybridity values recorded in our study were much higher than those reported previously (Blair et al., 2010b; Okii et al., 2014a). This might be explained by the fact that most of the accessions (>90%) were acquired from the National Gene Bank, which, in turn, had collected these accessions from subsistence farmers with a culture of keeping mixed seeds for consumption and subsequent planting seasons. Moreover, the higher allelic values of genomic than genic markers were comparable to those reported in some previous studies (Asfaw et al., 2009; Blair et al., 2010b; Okii et al., 2014b). The high genetic diversity of the Ethiopian common bean landraces was also evident when considering their ecological or geographical distribution. Such presence of high diversity in terms of both gene pools and the existence of ecologically- or geographically differentiation populations can have potential applications prospective common bean breeding programs in Ethiopia.

The presence of higher levels of gene flow within each gene pool than that found between gene pools observed in our study agrees with the result of Asfaw et al. (2009). This may be explained, in part, due to the lack of flowering synchronization, which could reduce inter-gene pool gene flow. A larger proportion of the accessions (i.e., 58%) were introgressions, which contradicts the report of Asfaw et al. (2009) about the lower level of introgression with Ethiopian and Kenyan bean landraces/cultivars. This, in turn, negates the assumption of the aforementioned authors implying that the genetic divergence in Ethiopian bean germplasm could be mainly due to the original differences in introduced germplasm from the primary centers of origin. Rather, the presence of a higher number of introgressions may be partially explained by the fact that the accessions were gene bank collections from farmers’ fields often characterized by a higher level of mixtures. The common practice of subsistence farmers in the country who cultivate for consumption and save segregant genotypes, resulting from any natural hybridization, as planting materials for subsequent generations, could result in such type of introgressions (Blair et al., 2010b; Worthington et al. 2012). A final noteworthy remark may be the fact that inter-gene pool introgressions are often endowed with useful combination of traits, including enhanced adaptation to environmental stresses, higher resistance to diseases and pests, and higher nutritional quality; hence, the introgressions identified in this study are of considerable importance in future bean breeding and conservation endeavors in Ethiopia. These merits of hybrids were evidenced in Islam et al. (2005) and Blair et al. (2010b), who reported that introgressions had higher mineral compositions than their respective non-hybrid parents. Consequently, it may be essential to tap into the useful genetic diversity found in such types of inter-gene pool introgressions, to be harnessed in further common bean breeding, improvement, and genetic conservation programs of beans in Ethiopia.

This study formulates new insights about the pattern and extent of genetic diversity and population structure of common bean landrace germplasm in Ethiopia. The results in the context of both the two gene pools of origin and ecological/geographic populations shed light on the presence of adequate genetic diversity organized into the Andean and Mesoamerican gene pools, and distributed across various ecological/geographic populations. This in turn should be strengthened by identifying the cluster groups identified by STRUCTURE via integrating molecular marker evaluations with phenotypic data.  

The authors have not declared any conflict of interest.

We are grateful to the financial and material support provided by the Rural Capacity Building Project (RCBP)-Ministry of Agriculture, Ethiopia; the African Biosciences Challenge Fund (ABCF) at the BecA-ILRI hub, Nairobi, Kenya; and Addis Ababa University. Moreover, we are thankful to the Ethiopian Biodiversity Institute, Ethiopia, and the National Common Bean Research Project based at Melkassa Agricultural Research Center, Ethiopia for availing the plant materials for our study.

For the molecular diversity assessment, total genomic DNA for each accession was isolated from a bulked leaf tissue sample of five randomly selected, one-week-old plants per accession using cetyltriethylammonium bromide (CTAB) method (Doyle and Doyle, 1990) with some minor modifications, as described in the following sections.

About 200 mg of fresh leaf tissue samples/leaf were placed in a 2 ml autoclaved and labeled Eppendorf tubes, covered by paraffin paper with a small slot at one side for air circulation, and freeze-dried for two days at-80⁰C. Subsequently, a drop of polyvinyl polypyrrolidone (PVPP) was added to the Eppendorf tubes. Then, 500 μl of 1× CTAB was added to each tube to break open cells and soluble cellular contents. Next, the contents in each tube were mixed using a Vortex, and kept in a gently-shaking water bath for 1 hour at 65°C. After the samples were taken out of the water bath, they were centrifuged at 14,000 rpm for 30 min, using an Eppendorf centrifuge (5417R). Afterwards, the supernatant suspension was transferred into new Eppendorf tubes, and 250 μl of potassium acetate was added. A total of 400 μl of ice-cold isopropanol was added to the supernatant solution harvested, after centrifuging the samples at 14,000 rpm for 30 min. At this point, the samples were left at -20°C overnight. The following day, the samples were removed from the -20°C freezer; centrifuged at 14,000 rpm for 30 min at -4°C. The supernatant was then poured off and the pellet dried. In order to remove the remaining isopropanol drops, the tubes were placed upside down on a paper towel. The pellets were air-dried for 30 min at room temperature.

Subsequently, 200 μl of TE and 3 μl of RNAse were added to each tube, which were then left in a water bath at 37°C. Following this, chlorophyll and some denatured proteins were removed by dissolving in a 200 μl mixture of phenol, chloroform, and isoamyl alcohol at a ratio of 25:24:1, which was mixed with manual inversions from 5 to 10 times. Next, the samples were incubated at room temperature for 10 min. Subsequently, a fixed volume of supernatant (180 μl) was harvested from each tube into new sets of 1.5 ml Eppendorf tubes. Three hundred μl of ice-cold 100% ethanol plus 15 μl of sodium acetate (at pH 5.2) was added to each. Following incubation at -80°C for 5 min, centrifugation was performed at 14,000 rpm for 30 min; the supernatant was poured off and the inside of each tube was washed with 200 μl 70% ethanol, and another centrifuging was applied at 14,000 rpm for 30 min at -4°C. Following this, DNA pellets were air-dried for an hour, and re-suspended with 30 μl of low salt buffer. DNA quality and quantity were measured by gel electrophoresis (using 1% agarose gel for 1 hour using λ-DNA as a size marker) (Figure 11).