Genetic Diversity in Banana and Plantains Cultivars from Eastern DRC and Tanzania Using SSR and Morphological Markers, Their Phylogenetic Classification and Principal Components Analyses

1.1. Background

Bananas and plantains are edible and vegetatively propagated parthenocarpic species of genus Musabelonging to the family Musaceaewhich according to Meng et al. [1] has wild seeded species native to South-East Asia. These seedless edible species are thought to have originated through intra- and interspecies crosses between M. acuminataColla and M. balbisianaColla, including some back crosses [2]. These species constitute a staple food, a key commercial crop and a major source of raw materials for both beverage and handicraft industries for hundred millions of people in the world. They include 20% of the population of the United Republic of Tanzania (URT), and its production promotes the country to be the second largest producer after Uganda in east Africa [3, 4, 5].

1.2. Problem statement and justification

The east African highland bananas (EAHBs) are currently threatened by several pests and diseases, which need diploid parents with farmers and other consumers’ desirable traits for inclusion in the breeding programme [6]. The edible diploid landrace ‘Mshale’ (Mchare [7], AA genomic group) of URT was identified to be highly similar to M. acuminataspp. malaccensiscv. ‘Pisang lilin’. Research using numerical taxonomy on AAA-EAHB genomic subgroup from Eastern DRC and Tanzania has shown certain level of relationship with ‘Mshale malembo’, suggesting that it is one of the ancestors [8, 9]. These observations were supported by Simmonds [7] and De Langhe et al. [10] but need to be confirmed at a molecular level. Such research has not yet been done and remains dearth for the inclusion of the Tanzania’s landrace in breeding programme. Elsewhere, such research using the AFLP technique has been conducted by Ude et al. [11], on phylogenetic origin of AAA-Gros Michel and AAA-Yangambi km 5. The technique has shown that these cultivars have similar ancestors that have contributed to their development. In this respect, M. acuminataspp. malaccensiscv. ‘Pisang lilin’ was identified as a source of one of their genomes (A). This supported the use of landrace AA ‘Paka’ from Zanzibar in the improvement of ‘Gros Michel’ in Jamaïca [7, 10].

1.3. Hypothesis, technology justification and objective

AFLP technique shows a dominant mode of inheritance and hence constitutes its limiting factor for this study. On the other hand, research using SSR markers has confirmed these preceding findings [12]. Moreover, the fact that the genetic map has 11 linkage groups of Pisang lilin was also reported [13]. Therefore, the determination of identity and confirmation of the contribution of ‘Mshale’ in the AAA-EAHB using microsatellite markers determined from Pisang lilin could be a useful tool for the regeneration of subgroups escaping genetic erosion due to pests. This would constitute different scientific point of view from the current belief that AAA-EAHB comes from somaclonal variation [14]. The study aimed to establish the cladistic relationship of the banana landrace ‘AA-Mshale’ in AAA-EAHB which may constitute a way for reconstituting the EAHB through breeding.

2.1. Plant materials

Cigar (unfurled) leaf samples from 25 accessions of bananas and plantains (Table 1) were collected from the existing banana gene bank in the Horticulture Unit of Sokoine University of Agriculture (SUA). The 25 accessions consisted of eight edible diploids (AA), nine, four and two triploids (AAA, AAB, ABB), two tetraploids (AAAA) genomic groups which were determined through numerical morpho-taxonomic classification [15]. Apart from the diploids and triploids AAA-EAHB subgroup, the other subgroups and genomic group were added as control to verify the accuracy of the ancestry. The SUA Musasp. germplasm was an in situfield conservation located in the plateau zone of Morogoro Urban District of Tanzania [5].

2.2. DNA extraction

The DNA of the 25 accessions (Table 1) was isolated using DNeasy Plant Mini Kit (Qiagen, USA; www.qiagen.com) following the manufacturer’s instructions, quantified in 2% agarose gel (in 0.5 TBE electrophoresis buffer) and stained in 5 μg/ml of ethidium bromide solution. The DNA quality was checked by ensuring that the 260/280-nm values ranged between 1.4 and 2.2 using spectrophotometer [12]. The PCR was performed using a Gene Amp PCR system 2700 thermocycler (Applied Biosystems). Each reaction was carried out in a total volume of 20 μl, containing 10 ng of genomic DNA, 1.2 mM MgCl₂, 10 mM dNTPs, 0.2 μM of each primer, 1.25 U of Taq polymerase and 10x Go Taq flex buffer (New England Biolabs, Inc.). Twenty-one SSR primer pairs (Table 2) distributed across the 11 linkage groups were used. This SSR primer selection was done among established linkage groups covering banana genome [13]. During amplifications, temperature cycling was conducted as follows: an initial denaturation step at 95°C for 5 min that was followed by 32 cycles of denaturation at 94°C for 1 min, annealing at each temperature as specified in Table 2 per primer pair for 1 min, and extension (elongation) for 90 s at 72°C. A final extension was carried out at 72°C for 7 min. For gel electrophoresis, a 10-μl aliquot of each amplification reaction was separated at 100 V for 2 h, using 2% agarose gels (0.5× TBE buffer). Gel images were photographed under UV illumination to check for amplicon size and PCR specificity. Allele sizes were estimated against 2-Log DNA Ladder molecular size standards. All samples were run with three replications starting from DNA extraction to maintain the integrity of the sample.

2.3. Data analysis

Alleles (0, 1, 2, …) were scored from 21 SSR marker pairs in the 25 accessions and were used to build the phenetic and cladistic trees. The data were analyzed using Numerical Taxonomy and/or Multivariate Analysis System package (NTSYSpc) version 2.1 (Exeter Software, Setauket, USA). The Manhattan method was used to assess similarity among the banana accessions. The genetic similarity matrices were then used to construct the dendrogram with unweighted pair group method with arithmetic mean (UPGMA) algorithms that employed the sequential, agglomerative, hierarchical and nested clustering procedure [16]. The cladistic kinship between accessions was determined based on neighbor joining coefficients using Dice dissimilarity coefficients (matrix using NTSYSpc 2.1. The scattered plot and accuracy of the trees were determined using principal component analysis (PCA) and cophenetic correlation method (from NTSYSpc 2.1). A two-way Mantel statistic test of 500 permutations was performed to get a cophenetic value.

3.1. Results

3.1.1. Molecular/genetic relatedness among accessions

The coefficient of dissimilarity varied from 0.28 to 0.66, being <1 or 100% showing no duplication among accessions from the 21 loci covering 11 linkage groups used as shown in Figure 1. Hence, the eight accessions belonging to AA-Mshale group were found to be genetically different. The dendrogram (Figure 1) established two main clusters (A and B). In the first cluster (A), AAA-Lujugira-Mutika accessions (‘Bukoba’/Musakala, ‘Muhowe’/Beer (Mbidde) and ‘Embwailuma’/Nakitembe) were clustered with the seven accessions of AA-Mshale (‘Ndyali’, ‘Mshale makyughu’, ‘Ilalyi’, ‘Mshale malembo’, ‘Nshonwa mshale’, ‘Ijihu’ and ‘Huti’). They included the tie of AAAA-FHIA (17 and 23) accessions with ‘Yangambi km 5’ (AAA-Ibota), ‘Mzungu mwekundu’ (AAA-Green-red) and ‘Green bell’ (AAA-Cavendish). Whereas in the second cluster (B), six heterogenomic accessions named ‘Kisukari’ (AAB-Silk), ‘Ngego I’, ‘Ngego Halisi’ and ‘Mzuzu’ (AAB-French Plantain), ‘Unyoya’ and ‘Bokoboko’ (ABB) were tied to three homogenomic accessions (AAA) ‘Jamaica’ (Gros Michel), ‘Kimalindi fupi’ (Dwarf-Cavendish) and Mwanjunjila (EAHB having a yellow male bud). The accession ‘King banana’ (AA) was an outline.

The genetic variation causes were allelic deletion or non-annealing and heterozygosis. The mMaCIR168 primer showed allele deletion in cultivars ‘Ndyali’ and ‘Mwanjunjila’ (first one and third three after (left) Ladder, Figure 2), and mMaCIR189 showed heterozygosis in cultivars ‘Mshale Makyughu’, ‘Ilalyi’ and ‘King banana’ (first six, nine and second three after ladder) while both primers showed a homozygote allele in cultivar ‘Mshale malembo’ (the first number six after the ladder). Similarly, alleles’ deletion (null alleles) was observed among 19 cultivars for primers mMaCIR117 and mMaCIR174. The alleles’ sizes resemble those of Hippolyte et al. [13].

The observed mutation has negatively influenced the principal component analysis (PCA) that resulted in poor fit of the clustering analyses with a cophenetic coefficient of 0.72 from distance matrix and 0.67 from product-moment correlation matrix. Consequently, the variation has spread over the principal component (PC) so that the three first PCs cannot hold the maximum of the variation (Figure 3) and hence weakened the value of PIC (Polymorphism Information Content).

3.1.2. Cladistic relationship

The cladogram showed three clades which revealed mono-, para- and polyphyly (A, B and C, Figure 4). The eight AA-Mshale accessions were subdivided into two clades. The first clade (A) was a monophyletic group composed of eight accessions in which six belonged to AA-Mshale genomic group (‘Ndyali’, ‘Mshale malembo’, ‘Ijihu’, ‘Nshonwa mshale’, ‘Huti’ and ‘King banana’) and two of triploid (‘Green bell’ (AAA-Cavendish) and ‘Mzungu mwekundu’ (AAA-Green-red).

The second clade (B, Figure 4) that encompassed AAA-EAHB accessions was subdivided into two subclades (B1 and B2) and formed paraphyletic group with the first clade. The first subclade (B1) was made of three accessions, ‘Mzuzu’, ‘Bukoba’ and ‘Yangambi km 5’, that belonged to AAB-French Plantain, AAA-EAHB-Musakala and AAA-Ibota, respectively. Whereas, in the second subclade (B2), the AAA-EAHB accessions ‘Muhowe’ and ‘Embwailuma’ shared the ancestry with AA-Mshale (Mshale makyughu and Ilalyi) and AAAA-FHIA (17 and 23). The last clade (C) had ‘Kimalindi fupi’ (AAA-Cavendish), ‘Mwanjunjila’ (AAA-EAHB) and Jamaica (AAA-Gros Michel) sharing a common ancestry with AAB-Silk (Kisukari), AAB-French plantain (Ngego Halisi and Ngego I) and ABB (Bokoboko and Unyoya). The clade (C) established a polyphyly with the two first clade (A and B) that had AA genomic group accessions. Whereas, in reference to accession ‘Jamaica’, there was a paraphyly between the clades B and C.

3.2. Discussion

This clustering from dissimilarity using UPGMA fairly confirms the relationship established by numerical taxonomy between the AA-Mshale malembo and the AAA-Lujugira-Mutika group determined by several authors [2, 7, 8, 9]. Likewise, the observed alleles’ differences among AA-Mshale accessions were in line with the morpho-taxonomic dissimilarity determined previously by the upcited authors. Moreover, the clone sets (Musakala, Nfuuka and Nakitembe) coined subjectively within the AAA-Lujugira-Mutika were linked with the different AA-Mshale accessions following their alleles’ closeness [16]. Interestingly, the clustering of AAA-Cavendish, AAA-Gros-Michel, AAA-Ibota, AAB-Plantain and AAB-Silk subgroups as sympatric is similar to results of [11, 12, 17], while they used other techniques or primers partly covering the 11 linkage groups [13]. This once more established the usefulness and reliability of the alleles from the 11 linkage groups in diversity and cladistic study.

The mono-, para- and polyphyletic relationships are in line with those revealed from numerical morpho-taxonomy [7, 8, 9, 10]. The para- and polyphyletic relationship may be explained by the hypothesis of back-crosses developed [2]. The back-crosses theory explains the role of the observed alleles deletion and rearrangement (heterozygosis) in the evolution of AA-Mshale malembo in the AAA-EAHB. These relationships were also similar to results from other microsatellites covering 10 linkage groups [17]. However, there is contrast with the statement of lack of convincing lineage between ‘Mutika-Lujugira’, ‘Red’, ‘Ibota’ and ‘Plantain’ subgroups, and the diploid M. acuminataaccessions. This may be explained by the poor fit of the clustering analysis and the spread of principal components over the variables due to observed mutation.