Agronomical and molecular characterization of banana germplasm

The objective of the present work was to characterize banana accessions from the Germplasm Bank at Embrapa Mandioca e Fruticultura Tropical (Brazil), using agronomical, physical and physicochemical characteristics of fruit and simple sequence repeats (SSR) markers. Twenty-six accessions were analyzed, in which high genetic variability was found, especially for the agronomical characters number of fruit and weight of bunch. Accessions with high contents of carotenoids (diploid 'Jaran'), polyphenols (triploid 'Caipira' and tetraploid 'Teparod') and vitamin C (diploid 'Tuugia' and an unknown triploid AAA) in the fruit were identified. Thirteen microsatellite primers revealed an average of 7.23 alleles, which showed high variability. A dendrogram was prepared using the Gower algorithm for the distance matrices obtained from the agronomical, physical and physicolchemical analysis of fruit and SSR markers. Adopting the average genetic divergence as the cut-off point, three clusters were found: G1, formed by the diploids 'Jaran', 028003-01 and M-48; G2, by the diploids 'Malbut' and 'Ido 110'; and G3, by 21 tri-and tetraploid accessions, including one diploid, 'Tuugia'. The triploids with the B genome 'Thap Maeo', 'Walha', 'Pacha Nadan' and 'Champa Madras' were grouped in G2. Results from this work can be used for breeding hybrids with good agronomical traits and fruit quality.


Introduction
Banana production is usually based on tri-and tetraploid cultivars; however, diploid genotypes are important sources of alleles for resistance or tolerance to biotic and abiotic factors. Banana breeding programs have generated promising tetraploid hybrids from crosses between triploids and improved wild diploids with important agronomic traits, including reduced height, disease and pest resistance and fruit quality (Silva et al., 2002).
Agronomical, physical and physicochemical characterization of fruit, allied with estimates of genetic variability assessed using molecular markers, is important for the selection of progenitors, in order to explore heterosis and develop better cultivars.
Many molecular markers, especially those associated with PCR (polymerase chain reaction) based methods, including RFLP (restriction fragment lenght polymorphism), RAPD (random amplified polymorphic DNA) and microsatellites or SSR (simple sequence repeats), have been broadly used to estimate genetic variability and phylogenetic studies in banana (Creste et al., 2004;Amorim et al., 2008;Amorim et al., 2009aAmorim et al., , 2009b. Of these, microsatellite technique has shown higher potential use in banana because it permits the detection of greater polymorphism and of co-dominant inheritance, and offers high reproducibility and easy interpretation. The objective of the present work was to characterize banana accessions from the Germplasm Bank at Embrapa Mandioca e Fruticultura Tropical using agronomical, physical and physicochemical characters of fruit and simple sequence repeats (SSR) markers.

Materials and Methods
Twenty-six accessions, from the Banana Germplasm Bank (BGB) at Embrapa Mandioca e Fruticultura Tropical, originated by tissue culture, including improved and wild diploids, triploids and tetraploids, were used (Table 2). Five plants per accession were evaluated in a complete randomized block design with a 3x2 m spacing. The plants were irrigated by aspersion, and the management was carried out according to technical recommendations for all genotypes (Alves & Oliveira, 1999). The soil was classified as a Latossolo Amarelo distrófico (Xanthic Haplustox). The climate type in Cruz das Almas is AW (Köppen classification), with an annual mean temperature of 24.5°C and rainfall of about 1,500 mm.
Yellow-Sigatoka was assayed under field conditions. Data was taken during the flowering period, following the methodology proposed by Stover (1972), according to the following grade scale: 0, without symptoms; 1, 1 to 10% of leaf blade with symptoms; 2, from 11 to 30% of leaf blade with symptoms; 3, from 31 to 50% of leaf blade with symptoms; 4, from 51 to 70% of leaf blade with symptoms; 5, more than 70% of leaf blade with symptoms.
The physical characters of fruit evaluated were: length; diameter; fruit weight; pulp weight; peel thickness; pulp diameter and pulp firmness. All analyses were done in triplicate.
Pulp samples from three fruits of each genotype, made up of a central piece and both extremities, were taken out for the physicochemical analyses. Pieces were shredded in a blender, adding water in a 1:2 ratio (pulp:water). The following variables were measured: soluble solids content (SS, ºBrix); pH and titratable acidity (TA), according to the Association of Official Analytical Chemists (1997); vitamin C (VIT, mg 100g -1 ), according to Terada et al. (1979); total carotenoids (CTN, µg g -1 ), according to Rodriguez-Amaya (1999); and total flavonoids (FLA, mg 100 g -1 ), as described by Rijke et al. (2006). Polyphenol extraction (PLF, mg 100 g -1 ) was done using a solution of 50% methanol and 70% acetone and quantification was performed with a spectrophotometer using the Folin-Ciocauteau reagent, according to the methodology proposed by Obanda & Owuor (1997). All analyses were done in triplicate.
For the molecular characterization, carried out at the Laboratorio de Virologia e Biologia Molecular of Embrapa Mandioca e Fruticultura Tropical, 13 microsatellite primer pairs were used: three belonging to the Ma series developed by Crouch et al. (1998), four from the AGMI series (Lagoda et al., 1998), five from the MaOCEN series (Creste et al., 2004), and a Mb 1-100 primer described by Oriero et al. (2006) (Table 1). Genomic DNA was extracted from young leaves using the CTAB method (Doyle & Doyle 1990).
Amplifications were carried out in the 9700 Perkin Elmer Thermocycler (Lincoln Center Drive, Foster City, USA), using the touchdown program: initial cycle of 3 min and 94ºC, followed by 40 s at 94ºC, 40 s at 55ºC, reducing one degree at each cycle, 1 min at 72ºC, totalizing 10 cycles, followed by 25 cycles of 40 s at 94ºC, 40 s at 45ºC and 60 s at 72ºC. Bands were separated in ultrapure 1000 agarose gel (Invitrogen Carlsbad, USA) at 3%, under standard conditions, and the amplification products were stained with ethidium bromide for allele visualization.
Fruit data were analyzed for variance, and the averages grouped by the Scott-Knott (1974) test at 5% of probability. The dissimilarity averages were calculated using the average Euclidian distance. The component analyses were done using the Singh (1981) criteria. All analyses were carried out using the Genes software package (Cruz & Schuster, 2004).
For SSR analyses, the amplified fragments were evaluated as absent (0) or present (1). Genetic similarity was calculated according to Nei & Li (1979).
A combined analysis using the matrices of the agronomical, physical and physicochemical data of fruit and the molecular (SSR) data was carried out using the Gower (1971) algorithm: in which:W ijk is the distribution of the k-i th variable for the total distance between individuals j and I; d ijk corresponds to the weight given to the ijk-i th comparison, with 1 for the valid comparisons and 0 for the invalid ones. This index was obtained by combining the binary data from the molecular markers and the quantitative data from the morphological characters, giving an estimation of a single similarity index varying from 0 to 1. The Gower distance matrix was used to cluster the 26 accessions by UPGMA (unweighted pair-group method averages), using the NTSYS-pc software (Rohlf, 2000).

Results and Discussion
The averages for all agronomical characters showed significant differences, except for the number of suckers during harvest (NS) ( Table 2). The coefficient of variation varied from 9.12 (PD) to 57.08% (NF), similarly to those observed by Amorim et al. (2009a) for the same variables.
Plant height (PH) varied from 1.44 m for 'Walha' to 3.54 m for 'Ambrosia', with an average of 2.79 m ( Table 2). Four clusters were formed with 028003-01 and 'Walha', classified in the last group with the lowest values by the Scott-Knott (1974) test.
There was great variability for plant height, which is a positive fact for breeding, since this make it possible to identify diploid progenitors for crosses aiming at the development of short hybrids.
The mean pseudostem diameter was 17.76 cm, with averages varying between 9.67cm for the diploid 'Idu-110' and 24.56 cm for the tetraploid 'Calipso' (Table 2). This trait is related to vigor and resistance to pseudostem breakage. Genotypes with thicker pseudostems are less susceptible to toppling over (Silva et al., 2002).
The highest value observed for live leaves during harvest was 9.20 for the Tropical and Maravilha cultivars, and the smallest for the Towoolle, with 2 leaves. Fruit filling is directly correlated with number of live leaves during harvest. According to Soto Ballestero (1992), cultivars from the Cavendish subgroup generally need at least eight active leaves per plant for good fruit development.
The shortest mean stalk length was observed in 'Tuugia', with the highest value for 'Champa Madras'. There was a direct relationship between stalk diameter and stalk weight, which had the highest values in 'Ambrosia' and 'Calipso' (Table 2), and the smallest in 'Tuugia'. The mean number of hands and fruit per bunch was 6 and 83, respectively ( Table 2). The highest numbers of fruit were found in 'Thap Maeo', 'Ambrosia', 'Jaran' and 'Calipso'. These two traits are important in banana genetic studies, since the hand constitutes the market unit, and the increase in the number of hands can increase bunch weight, a trait that expresses genotype yield (Silva, 2002).
Five clusters were formed for resistance to yellow Sigatoka during flowering. Most genotypes were resistant, except 'Jaran' and 'Malbut' ( Table 2). All the physical and chemical characters of fruit, except titratable acidity, showed significant differences ( Table 3). The variation coefficients varied from 0.86% for flavonols to 31.24% for fruit weight.
The mean fruit length was 13.31 cm, with averages varying from 6.78 cm for 'Jaran' to 18.7 for 'Calipso', forming five groups. In general, tetraploids had higher averages than triploids or diploids. Similar behavior was observed for the diameter and average weight of fruit and for pulp weight and diameter, which formed five, four, four and six groups, respectively. The average peel thickness was 0.24 cm, varying from 0.13 for the diploid 'Idu-110' to 0.47 for the triploid 'Bakar'. Through the Scott-Knott (1974) test, three groups were formed (Table 3). The same behavior occurred for pulp firmness (PF), with averages varying from 0.67 Lb ('M-48') to 1.2 Lb ('Terapod').
The average for soluble solids was 19.48 ºBrix, varying from 14.60 ºBrix for 'Towoolle' to 25.70 ºBrix for 'Teparod'. Three groups were formed by the Scott-Knott clustering test. These results are in agreement with Soto Ballestero (1992).
The highest vitamin C content was found in 'Teparod' and in the unknown AAA triploid. Results reported in the present work are in agreement with those of Amorim et al. (2007).
Regarding molecular analysis, 94 alleles were found, with an average of 7.23 alleles per primer. The largest number of alleles was identified for primer AGMI 103-103 (11), and the smallest for MAOCEN 01 (Table 1). The average number of alleles per SSR locus is similar to other reports for banana (Amorim et al., 2008;Amorim et al., 2009a).
The polymorphism information content (PIC) varied from 0.72, for AGMI 24-25, MAOCEN 03 and Mb 1-100 primers, to 0.91 for the MAOCEN 12 primer, with an average of 0.81. The correlation between the number of alleles and the PIC value was high (r = 0.76, p ≤ 0.005), indicating that the primers with larger number of alleles had greater discriminatory power.
Re-sampling indicated that 85 alleles were enough for a precise estimation of the genetic divergence between the accessions. Correlation between the matrix considering 80 alleles and the matrix with 94 alleles was 0.98, with deviated squares (SQ d ) of 0.27 Table 3. Average of physical and physicachemicol fruit characters, evaluated for the 26 banana accessions from the Embrapa Mandioca e Fruticultura Tropical Germplasm Bank (1) .
(3) Unknown triploid. ns Nonsignificant.*Significant at 5% probability. and stress value (E) of 0.0475. According to Kruskal (1964), a value of E≤ 0.05 is indicative of excellent precision of the estimates. Similar results have been reported by other authors for banana (Creste et al., 2004;Amorim et al., 2008).
The estimates of the relative contribution of each variable, for expressing genetic diversity using Singh´s (1981) criteria, indicated that for the agronomic variables, the number of fruit contributed with 83.80% of the variation (Table 5), while fruit weight and pulp weight contributed with 59.42% and 30.32% of the variation, respectively, to the fruit variables (Table 5).
Considering the distance matrices for the agronomic, fruit and molecular (SSR) data, a dendrogram was constructed using the Gower (1971) algorithm ( Figure 2). The cophenetic value was high (r = 0.88, p<0.0001, 10.000 permutations) and adequate, since rt 0.56 is considered ideal, reflecting  a good fit with genetic similarity (Vaz Patto et al., 2004).
Adopting the average genetic divergence (0.57) as the cut-off point, three groups were formed: G1, formed by the diploids 'Jaran', 028003-01 and M-48; G2, formed by the diploids 'Malbut' and 'Ido 110'; and G3, formed by 21 tri-and tetraploid, including one diploid 'Tuugia' accession. The triploids with the B genome, 'Thap Maeo', 'Walha', 'Pacha Nadan' and 'Champa Madras' were grouped in G2. The results from this work can be used as sources for banana genetic breeding, for the development of tri-and tetraploid hybrids with good agronomic traits.
The most divergent accessions 'Porp' (tetraploid AAAB) and 'Jaran' (diploid AA) also showed divergent concentrations of functional compounds (carotenoids, flavonoids, polyphenols and vitamin C) (Table 3). These accessions can be intercrossed to develop segregating populations for genetic mapping and to identify genes associated with these compounds. Furthermore, other parental combinations can be set up for the development of hybrids with promising agronomic traits and potential use as functional foods.
S.j: contribution of the x variable for the Euclidian distance among i and i' genotypes. Table 5. Relative importance (S.j.) of agronomic and fruit characters for the study of genetic diversity, in 26 banana accessions from the Embrapa Mandioca e Fruticultura Tropical Germplasm Bank.