Genetic diversity in table grapes based on RAPD and microsatellite markers

The objective of this work was to analyze the genetic diversity of 47 table grape accessions, from the grapevine germplasm bank of Embrapa Semiárido, using 20 RAPD and seven microsatellite markers. Genetic distances between pairs of accessions were obtained based on Jaccard's similarity index for RAPD data and on the arithmetic complement of the weighted index for microsatellite data. The groups were formed according to the Tocher's cluster analysis and to the unweighted pair-group method with arithmetic mean (UPGMA). The microsatellite markers were more efficient than the RAPD ones in the identification of genetic relationships. Information on the genetic distance, based on molecular characteristics and coupled with the cultivar agronomic performance, allowed for the recommendation of parents for crossings, in order to obtain superior hybrids in segregating populations for the table grape breeding program of Embrapa Semiárido.


Introduction
The development of new grape cultivars adapted to subtropical and tropical conditions has been a major goal for grape breeding programs in Brazil (Ritschel et al., 2010).In Northeastern Brazil, the main objective of recent grape breeding projects has been the development of seedless table grapes, which are more resistant to diseases and well-adapted to the warm and dry environment of the Brazilian semi-arid region.
The knowledge of the genetic divergence can help breeders to concentrate efforts on the most promising combinations.The heterosis expressed in crosses is directly related to the genetic divergence between parents.The feasibility of studying genetic divergence as a criterion for selecting parents for hybridization has been demonstrated for several species, among them grapevine (Oliveira et al., 2005).
The objective of this work was to analyse the genetic diversity of 47 table grape accessions, from the grapevine germplasm bank of Embrapa Semiárido, using 20 RAPD and seven microsatellite markers, for selecting parents with higher heterotic potential, allowing for crosses and the development of new seedless table grape cultivars adapted to the environmental conditions of Northeastern Brazil.

Materials and Methods
Forty-seven table grape accessions from the grapevine germplasm bank of Embrapa Semiárido, Juazeiro, BA, Brazil, were analyzed (Table 1).The cultivars were selected for their desirable agronomic traits and potential for use as parents in a table grape breeding program.
Genomic DNA was extracted using a modified CTAB protocol (Lodhi et al., 1994).In the final step, DNA pellets were suspended in 100 µL of 1X Tris-EDTA buffer, and stored at -20 o C. DNA quality was assessed on agarose gel 0.8% (w/v) stained with ethidium bromide solution (10 mg mL -1 ).
One hundred and forty RAPD primers (Operon Technologies, Alameda, CA, USA) were initially screened for six grape cultivars.Twenty primers which produced reproducible and polymorphic bands were used to amplify all cultivars.Polymerase chain reaction (PCR) amplification was performed in 12.5 µL final volume containing 10 ng µL -1 template DNA, 1.25 µL 10X PCR buffer (Invitrogen Co., Carlsbad, CA, USA), 25 mmol L -1 MgCl 2 (Invitrogen), 2.5 mmol L -1 of each dNTP, 4 mmol L -1 of each primer, 10 mg mL -1 BSA (bovine serum albumin) and 1 unit Taq DNA polymerase (Invitrogen).Amplification was done in a MJ Research PTC-100 thermal cycler (GMI Inc., Ramsey, Minnesota, USA), programmed for two cycles of 1 min at 94ºC, 30 s at 35ºC, and 1 min at 7ºC, followed by 40 cycles of 15 s at 94ºC, 30 s at 35ºC, 1 min at 72ºC, and a final extension step of 7 min at 72ºC.Amplification products were resolved by electrophoresis on 1.5% agarose gels run at 100 V for 3 hours in 1X TAE buffer containing ethidium bromide solution (10 mg mL -1 ).
Seven well-characterized, highly polymorphic grapevine markers with simple sequence repeat (SSR) were used: VVS2 (Thomas & Scott, 1993); VVMD5, VVMD7, VVMD27, VVMD31 (Bowers et al., 1996(Bowers et al., , 1999)); VrZAG79 and VrZAG62 (Sefc et al., 1999).One primer in each primer pair was 5'-labelled with one of the following fluorescent dyes: 6-FAM, HEX and NED.The PCR amplification was done separately for each locus in 10 µL final volume containing 2.5 ng µL -1 template DNA, 10 pmol of each primer, 2.5 mmol L -1 of each dNTP (Applied Biosystems, Foster City, CA, USA), 1 µL 10X Gold PCR buffer (Applied Biosystems), 2 mmol L -1 MgCl 2 (Applied Biosystems) and 0.5 units AmpliTaq Gold DNA polymerase (Applied Biosystems).The PCR reactions were carried out using a MJ Research PTC-100 thermal cycler.The cycling program for all markers consisted of the following steps: 5 min at 95°C, followed by 35 cycles of 30 s at 95°C, 45 s at 60 o C, 1 min at 72 o C, and a final extension step of 7 min at 72 o C. Amplifications were checked by running an aliquot of 4 µL of the PCR reaction product on 2% ethidium-bromide-stained agarose gels.Aliquots (1.7 µL) of PCR products were mixed with 0.7 µL of formamide, 0.4 µL of dye and 0.2 µL of DNA size standard Genescan-500 ROX (Applied Biosystems).Samples were denatured at 94°C for 2 min, and loaded onto an ABI 377 DNA sequencer (Applied Biosystems).To obtain consistent and reliable results, each sample was amplified and analyzed twice at each marker.Four grape cultivars with well-established SSR profiles (Chardonnay, Carignane, Riesling and Thompson Seedless) were loaded on each gel for consistent scoring from one gel to another.
RAPD bands were scored for their presence (1) or absence (0), and the resulting data matrices were analyzed using the Genes 2007.0.0 software (Cruz, 2008).Similarity matrices were computed with the Jaccard's coefficient.
SSR fingerprints were detected with GeneScan analysis software v. 3.1, and the alleles were scored using the Genotyper DNA fragment analysis software v. 2.5.2.(Applied Biosystems).The similarity index between two individuals was estimated by the arithmetic complement of the weighted index.
A dendrogram was constructed by cluster analysis based on the UPGMA algorithm.The input data were also processed by cluster analysis, using the Tocher's optimization method (Cruz et al., 2004).

Results and Discussion
Out of the 140 RAPD primers tested, only 20 produced reproducible and polymorphic bands.One hundred and eleven loci were identified, which represented 81.6% of total amplified bands.The number of recordable loci for each primer ranged from 1 to 10 with an average of 5.5.The size of amplification products ranged from 380 to 3,000 pb.
The polymorphism level found in the present work was higher than in previous studies on grapevine (Luo & He, 2001;Narvaéz et al., 2001;Borrego et al., 2002;Ulanovsky et al., 2002).According to Tamhankar et al. (2001), the polymorphism levels depend on the analyzed species; these authors obtained 94% of polymorphism for wild grapevine species and rootstocks, and more than 90% for Vitis vinifera genotypes; however, almost all bands were monomorphic among Vitis labrusca.The patterns obtained by RAPD analysis are primer and sample-dependent (Luo & He, 2001).The number of RAPD bands scored is also important to obtain accurate estimates of genetic relationships.The clustering in the dendrogram is completely rearranged if the number of scored bands is 100-150.The estimation of genetic distances improved with the progressive increase of the number of bands (Fanizza et al., 2000).The number of scored bands in the present study is in accordance with Narváez et al. (2001), Pinto-Carnide et al. (2003), and Kocsis et al. (2005).
The values of genetic distance based on the Jaccard's coefficient ranged from 0.25, between the intravarietal 'Itália' clones 'Italia Muscat' and 'Brasil', which were the most closely related accessions, to 0.68 between 'Piratininga' and 'Niagara Rosada', the most distantly related accessions.The closest genetic relationships among 'Italia' clones ('Italia Muscat', 'Benitaka' and 'Brasil') had been expected and are in accordance with previous works in Brazil for the cultivar Italia and its red bud sports (Leão et al., 2009).
The genetic similarity coefficient, used for microsatellite data, was the arithmetic complement of the weighted index, with values of genetic distances ranging from 0 to 1, in which 0 is the coincidence of all alleles and indicates that the pair of accessions corresponds to a same genotype, and 1 is the pair of accessions that shares no alleles and shows maximum genetic distance.
The 20 RAPD primers analyzed in the present study could distinguish all accessions.However, the results based on seven SSR loci showed coincidence of allelic profiles for three pairs of accessions -'Thompson Seedless' and 'Canner', 'A1581' and 'A1105', and 'Centennial Seedless' and 'Dawn Seedless' -, suggesting a close genetic relationship between them.The number of primers sufficient for reliable varietal identification depends on the nature and discriminating power of each primer (Tessier et al., 1999).Normally, six primer pairs are sufficient for differentiating between genotypes (This et al., 2004), but closely related cultivars require a larger number of primers (Meredith et al., 1999).In the present study, the seven SSR loci recommended by the GEN-RE081 project were not sufficient for differentiating the 47 genotypes obtained.
The clusters and dendrograms, regardless of the multivariate analysis method, showed variations depending on the molecular marker used, which is in agreement with Merdinoglu et al. (2000), who mention that, although RAPD, SSR and amplified fragment length polymorphism (AFLP) have been able to distinguish seven groups of grape cultivars, the topology of the dendrograms was specific for each marker, and that SSR better reflects the genetic relationships between groups and their geographical origin.According to Pinto-Carnide et al. ( 2003), both molecular markers were able to identify Portuguese grape varieties, but the advantage of microsatellites is that they allow for comparison between different laboratories.Tocher's cluster analysis applied to RAPD data showed 15 groups (Table 1).The main group concentrated 54.2% of the accessions and was subdivided into 10 subgroups.RAPD was efficient to distinguish V. vinifera L. and V. labrusca L. species.'Italia Muscat', 'Brasil' and 'Benitaka' are bud sport clones of 'Italia'; they were in the same cluster and also in the same subgroup, which confirmed the closest genetic distances obtained among them.Tamhankar et al. (2001) also observed the clustering of bud sport selections from 'Thompson Seedless' grown in India.
The Tocher's cluster analysis based on microsatellites resulted in six groups.There was a main group with 76.6% of the accessions (Table 3), which was divided into 11 subgroups.
A dendrogram was generated based on the UPGMA analysis using RAPD markers and could distinguish 11 groups when a relative genetic distance of 86.6% at the cut point was considered.The first main group was composed of 30 accessions (63.8%).The second one included six accessions: 'Venus', 'Blush Seedless', 'A1105', 'A1118', 'Cardinal' and'Beni Fuji', showing similarity to group 3 obtained by Tocher's method.'Flame Seedless' and 'Juliana' were joined in the fourth group.V. labrusca cultivars, 'Isabella Precoce' and 'Niagara Rosada', similarly to which occurred using Tocher's cluster analysis, were together in another group, with 0.40 of genetic similarity.The 'BRS Clara', 'Beauty Seedless', 'BRS Morena', 'Christmas Rose', 'Maria' and 'Piratininga' accessions were each separated in a distinct group, because they were distantly related to the rest of the accessions (Figure 1).
The first major group was subdivided in 12 subgroups with a genetic distance of 73.6% at the cut point.The dendrogram was correlated with the genealogy of accessions.Subgroup 1 was composed by 'Italia' clones -'Italia Muscat', 'Brasil' and 'Benitaka' -and 'Moscatel Nazareno'. 'Moscatel Nazareno' and 'Italia' have 'Muscat Hamburg' as their common parents.Another subgroup was formed by 'Crimson Seedless' and 'Princess', the first being a parent of the second.'Red Globe' and 'Ruby Seedless' formed a third subgroup; both are descendants of the cultivar Emperor.However, the descendants of 'Thompson seedless' -'Canner', 'Fiesta', 'Lakemont Seedless' and 'Perlette' -were separated into distinct subgroups.
In the dendrogram generated from the matrix of pairwise distances, based on microsatellite data, five main groups were distinguished with a genetic distance of 84% at the cut point (Figure 2).The first group was the largest, containing 38 accessions or 80.8% of all the analyzed accessions.The second group was similar to group 4 obtained by Tocher's method, and was composed of 'Dattier Beyrouth', 'Princess' and 'Feal'.The French hybrids 'Seyve Villard 12375' and 'Seyve Villard 20365' formed a third group, coinciding with the Tocher group 2. 'Niagara Rosada', 'Isabella Precoce' and 'Venus' were included in the fourth group.The largest genetic distance was obtained for the cultivar Kyoho (d = 0.897), isolated in group 5.The UPGMA method allowed for clustering of the 'Niagara Rosada' and 'Isabella Precoce' (both from V. labrusca) in the same group.This was not observed when the data were analyzed by Tocher's method.'Venus' is in the same group; it is an interspecific hybrid from the University of Arkansas and has also V. labrusca in its genetic background.
The first group was subdivided in nine subgroups with a genetic distance of 66.76% at the cut point.Subgroup 1 in this dendrogram corresponded to Tocher subgroup 2. Interestingly, all accessions in this group Figure 1.Cluster analysis of 47 table grape accessions by the UPGMA clustering using Jaccard's dissimilarity index, based on 111 RAPD molecular markers.The four main groups are identified as G1, G2, G3 and G4.
(except for 'Blush Seedless') are descendants from 'Thompson Seedless'.'Blush Seedless' showed no relationships with the other accessions, and, therefore, did not share alleles with them, which shows a possible misnaming or a genotyping error.
Subgroup 2 coincided with subgroup 1 by Tocher's method, including selections coming from the grape breeding program of the University of Arkansas ('A1105', 'A1581' and 'A1118').Subgroup 3 was similar to subgroup 5, obtained by Tocher's method, and was composed of cultivars which have 'Emperor' as a common parent (except 'BRS Linda').Subgroup 4, formed by cultivars Dona Maria, Muscat of Alexandria and Estevão Marinho, was the same as subgroup 8 by Tocher's method.'Muscat of Alexandria' is a parent of the Portuguese 'Dona Maria'.
'Centennial Seedless', 'Dawn Seedless', 'BRS Clara', 'Branca Salitre' and 'BRS Morena' formed the fifth subgroup and differed from the subgroup 3 of Tocher's method due to the inclusion of 'BRS Morena'.However, Figure 2. Cluster analysis of 47 table grape accessions by the UPGMA clustering, using the arithmetic complement of weighted index based on seven microsatellite markers.The four main groups are identified as G1, G2, G3 and G4.
The genealogy of 'Ferlongo' and 'Cardinal' has 'Alphonse Lavalle' as a common parent.They were included in subgroup 6 with 'Moscatel Nazareno'.
Subgroup 7 was composed by six cultivars: 'July', 'Red Globe', 'Piratininga', 'Beni Fuji', 'Perlette' and 'Muscat Hamburg', and was similar to the subgroup of Tocher's analysis.In this subgroup, 'July Muscat', 'Piratininga ' and 'Beni Fuji' share 'Muscat Hamburg' in their genetic background. 'Brasil', 'Italia Muscat' and 'Benitaka', bud sports of cultivar Italia, and the cultivar Juliana were included in subgroup 8. 'Juliana' shared 50% of alleles with 'Italia Muscat', since 'Italia' is one of its ancestors.The Brazilian accession 'Patricia', was separated into subgroup 9.In the present study, microsatellites could distinguish Brazilian intravarietal clones.According to Cretazzo et al. (2010), in a similar study in Spain, the most efficient technique to distinguish clones within autochthonous cultivars was M-AFLP.
Both cluster analysis methods showed a good agreement in the formation of groups; however, UPGMA resulted in a greater consistency between groups and cultivar genealogies.Microsatellites were more efficient than RAPD to study genetic relationships.Thus, the matrix of pairwise distances based on microsatellites was used to establish the five least related accessions (higher genetic distances) for ten seedless table grapes, which are important for their agronomic and commercial traits.Knowledge of the most divergent accessions is essential to guide the breeder in choosing the best combinations.However, other aspects should be considered to select parents for hybridization in addition to genetic divergence, such as the agronomic characteristics of cultivars, as well as breeding program goals.
In a table grape breeding program, whose main objective is the development of seedless grapes, the selection of parents should be based on the results shown in Table 4.For the development of new seedless grape cultivars, the strategy adopted by breeding programs worldwide has been the cross between two seedless grapes, followed by embryo rescue, which may increase the seedless trait frequency in the progenies  et al., 1998).Therefore, crossings could be recommended between 'A Dona' and 'Centennial Seedless' and 'Princess', using the embryo rescue technique.Still, aiming to incorporate resistance to fungal diseases, 'Isabella Precoce' should be used for its moderate resistance to bacterial canker and may also be used to obtain interspecific hybrids in crosses with 'BRS Clara', 'BRS Linda', 'Thompson Seedless' and 'Perlette'.The cross 'Thompson Seedless' x 'Isabella Precoce' can still be a good option to obtain mapping populations, in the search for genes linked to seedless characteristic and to resistance to bacterial canker trait.

Conclusions
1. RAPD markers are sufficient to distinguish all table grape accessions analyzed; however, microsatellites are only able to distinguish cultivars.
2. The cluster analyses using Tocher and UPGMA methods have a good agreement in group formation, and UPGMA provides a better consistency between groups and genealogy.
3. There was no relation between the dissimilarity matrix, formation of groups and topology of dendrograms, when RAPD and microsatellite data were used.
4. Assessment of genetic diversity of a subset of table grape cultivars, in addition to the knowledge of their agronomic characteristics, allow for the recommendation of crosses between seedless grapes, which can contribute to the table grape breeding program in Northeastern Brazil.

Table 1 .
Forty-seven table grape accessions (name, geographical origin, species and pedigree), analyzed by RAPD and SSR markers according to the Vitis International Variety Catalogue (VIVC, 2011).
(Cardinal x Thompson)x (Red Malaga x Tefafini Ahmer) x (Muscat of Alexandria x Thompson)

Table 2 .
Number of alleles, allele size range, expected heterozygosity (He), polymorphic information content (PIC) and maximum frequency of allele of the seven loci used for SSR analysis of 47 table grape accessions.

Table 3 .
Clustering according to Tocher's method, for 47 table grape accessions based on 111 RAPD and seven SSR loci.

Table 4 .
Cultivars with the highest genetic distances from ten seedless table grape accessions, based on the arithmetic complement of the weighted index for SSR data.