Abstract
Grapevine rootstocks are a complex group of plants, most of them hybrids derived from native North American Vitis species that are used to provide resistance against phylloxera and soilborne problems. A representative group of rootstock accessions and cultivars of the Vitis species commonly used in rootstock breeding (V. vinifera, V. berlandieri, V. riparia, and V. rupestris) and conserved in the largest European germplasm banks of Vitis were analyzed using sequence tagged microsatellite sites (STMS) and amplified fragment length polymorphism (AFLP) markers. The STMS analysis allowed assigning a microsatellite genotype to most of root-stock cultivars, although it revealed numerous misclassified accessions in the studied collections. Genetic similarity among the different genotypes was analyzed using AFLP, which provided information on the genetic relationships within and between hybrid groups.
Grapevine rootstocks represent one of the most effective long-term uses of a biological control mechanism for an agricultural pest. They were introduced to viticulture at the end of the 19th century to fight grape phylloxera (Daktulosphaira vitifoliae Fitch). Natural resistances to this insect pest did not exist in cultivated European grapevine (Vitis vinifera L.), but could be found in related species from North America, where grape phylloxera originated. Different hybrid cultivars, derived from crosses between American Vitis species (such as V. berlandieri Planchon, V. riparia Michaux, V. rupestris Scheele) or between American Vitis and a few cultivars of V. vinifera, were produced and tested as rootstocks. Acceptance of this new plant material depended on the combination of good levels of resistance to phylloxera and compatibility and conservation of fruit-quality traits of the grafted cultivars. There are no estimates for the exact number of root-stock cultivars that exist today, but each one represents a unique genotype that is maintained through vegetative propagation. Each rootstock is characterized by a specific behavior regarding adaptation to soil type, compatibility with V. vinifera cultivars, and resistance to soilborne pest and diseases (Read and Gu 2003). However, the number of rootstock cultivars currently used worldwide is very low because the grape industry relies on a few that are well-characterized and historically dependable. For example, 90% of Italian vineyards are grafted using only seven rootstock cultivars, while only six are used in France and Spain (Javier Provedo 2004, personal communication). There are, however, many more rootstock cultivars that are conserved in situ in germplasm banks. This material is a valuable source of genetic variation for possible use in breeding programs.
Specific genetic diversity and initial efforts to identify rootstock cultivars were based on ampelographic traits (Galet 1956, Ravaz 1902), although the results of these traditional descriptive methods may vary depending on environmental conditions. Differentiation of closely related material is difficult because the original genetic pool used to generate hybrid rootstocks was narrow and few detailed descriptions are available. Furthermore, after rootstocks are grafted with fruiting varieties, it is not possible to observe their leaves for ampelographic identification. Molecular markers offer a wide choice of analytical systems to overcome these difficulties by analyzing and comparing differences in DNA sequences. They have been widely used to analyze diversity in many cultivated crops, including grapevine (Thomas et al. 1994). Sequence tagged microsatellite sites (STMS) are highly polymorphic molecular markers in Vitis and have become the markers of choice for genetic identification because of their codominant behavior and ease of data exchange. They have demonstrated their usefulness in cultivar characterization and identification of V. vinifera, allowing detection of homonyms and synonyms (Ibañez et al. 2003), resolution of identification problems (Lopes et al. 1999), and rootstock characterization (Lin and Walker 1998, Sefc et al. 1998). However, because of their monogenic behavior, multiple STMS must be analyzed to prove identity between accessions. On the other hand, high throughput molecular markers, such as amplified fragment length polymorphism (AFLP) (Vos et al. 1995), yield many fragments per reaction, providing a rapid view of genetic relationships among uncharacterized materials. They have been successfully applied to study intraspecific and interspecific genetic relationships in many different plant species and cultivars (Cervera et al. 2005).
There is relatively little information on the genetic diversity of grape rootstocks. As a first approach to characterize the genetic diversity of rootstocks, STMS and AFLP markers were used to genotype accessions representative of some of the main rootstock cultivars, including the most frequently used rootstock cultivars in Europe and samples of different Vitis species involved in their published pedigrees. This study focused on the grape germ-plasm collection at El Encín, but samples from other collections were included to complete the characterization and to confirm the results. Nine microsatellite loci were used to characterize these collections, searching for redundancies or misclassified accessions. This analysis revealed that several accessions provided by different germ-plasm banks had different genotypes. In addition, a few accessions representing different rootstocks shared the same genotype. A microsatellite genotype was finally assigned to most of rootstock cultivars after rejection of misclassified material. The genetic relationships among most of the accessions were also analyzed using AFLP, providing information about the existing variation between and within each hybrid group.
Materials and Methods
A total of 208 samples from different germplasm banks were studied. They corresponded to 83 accessions, including Vitis species originally used in rootstock breeding programs (V. vinifera L., V. berlandieri Planch., V. riparia Michx., and V. rupestris Scheele) and the most relevant interspecific hybrid rootstocks derived from crosses between them, representing rootstocks commonly used in Europe (Table 1⇓). Many samples, 133 plants corresponding to 66 accessions, were from the Germplasm Bank of Vitis of El Encín. Sixty-six additional accessions from other European grape collections were included to confirm genotypes and pedigree information. DNA samples were extracted from leaves, which were harvested, immersed into liquid nitrogen, and stored at −80°C until use, with a DNeasy Plant Mini Kit (Qiagen, Hilden, German).
STMS analysis.
Each sample was genotyped using nine different nuclear STMS loci: VVS1, VVS2, VVS5, and VVS29 (Thomas et al. 1993); VVMD5 and VVMD7 (Bowers et al. 1996); and ssrVrZAG47, ssrVrZAG62, and ssrVrZAG 79 (Sefc et al. 1999). These markers were chosen based on their ability to reveal high levels of polymorphism in previous studies (Lin and Walker 1998, Sefc et al. 1998, Thomas et al. 1994). To identify possible mistakes, two independent plants from the El Encín germplasm bank were analyzed for each accession. STMS analysis was performed using multiplex PCR protocols following the method of (Ibáñez 2000), in which successive reactions amplify sets of three microsatellites: set A (VVS29, VVS5, and VVMD7), set B (VVS1, VVS2, and VVMD5), and set C (ssrVrZAG47, ssrVrZAG62, and ssrVrZAG79). PCR reactions were carried out using the forward primers labeled with fluorochromes to detect amplified products using an ABI-Prism 310 sequencer and GeneScan software (version 3.1; Applied Biosystems, Foster City, CA). The analysis was repeated for several samples at locus VVS1 using Deep Vent DNA Polymerase (New England Biolabs, Ipswich, MA), which has 3′ to 5′ exonuclease activity, to avoid +A fragments, because of the presence of alleles differing in one base pair. PCR amplification reaction was carried out in a total volume of 25 μL of Thermopol reaction buffer [10 mM KCl, 10 mM (NH4)2SO4, 20 Mm Tris-HCl (pH 8.8), 0.1% Triton X-100, 2 mM MgSO4], 2.5 mM MgSO4, 400 μM of each dNTP, 0.4 μM of VVS1 primers, 50 ng of DNA, and 0.25 U of Deep Vent DNA Polymerase. PCR was performed at 94°C for 5 min, 35 cycles at 94°C for 1 min, 55°C for 1 min, and 72°C for 90 sec.
Allele sizes were adjusted to an exact number of bases using the algorithm described elsewhere (Ghosh et al. 1997). Plants showing a single peak for a locus were considered as homozygous for that locus, with the only exception of locus VVS5, for which the existence of null alleles is known (Thomas et al. 1994). The effective number of alleles (Ne) was calculated according to Kimura and Crow (1964). Observed and expected heterozygosity (He = 1−∑pi2) (Nei 1973), probability of identity (PI = ∑pi4+∑[2pi x pj]2) (Paetkau et al. 1995), and probability of null alleles (r = [He−Ho]/[1+He]) (Brookfield 1996) were calculated using IDENTITY (version 1.0; Center for Applied Genetics, University of Agricultural Sciences, Vienna).
AFLP analysis.
A subset of 65 samples (Table 1⇑), including accessions of commonly used rootstock cultivars, crosses used for rootstock breeding, and representative accessions of Vitis species used for rootstock breeding, were subjected to AFLP analysis. Duplicated material was reduced to one entry for this analysis. When more than one microsatellite genotype was found for an accession, we only used the sample considered as “true-to-type” based on STMS results. AFLP analysis was performed following a previously described protocol (Vos et al. 1995) with minor modifications (Cervera et al. 1998). Three different primer combinations were used: combination I [2EcoRI (+ACC, +ACT)/MseI +CAT], combination II [2EcoRI (+ACC, +ACT)/MseI +CTG], and combination III [EcoRI +ACC/MseI +CTT]. AFLP gels were visually scored by two persons and only polymorphic, distinct, reproducible, and well-resolved fragments were used to build a binary matrix of presence/absence. A graphical representation of genetic similarities between pairs of accessions based on Dice coefficient (Dice 1945) and UPGMA clustering method (unweighted pair-group method with arithmetic averaging) (Sneath and Sokal 1973) was obtained using the program NTSYSpc (version 2.0; Exeter Software, Setauket, NY).
Results
Microsatellite analysis.
All samples were genotyped at nine STMS loci (Table 1⇑), with genetic parameters estimated for each locus (Table 2⇓). The STMS loci showed a high number of alleles, ranging from 9 (VVS29) to 23 (ssrVrZAG47), with an average of 17.1 per locus and a total of 154 alleles. Two loci displayed predominant alleles: VVS1, bearing the 190bp allele with a frequency of 0.47, and VVS29, bearing the 172bp allele with a frequency of 0.61. The number of genotypic combinations found for each locus ranged from 17 (VVS29) to 62 (ssrVrZAG47), corresponding to low PI values (probability of identical genotypes) (Paetkau et al. 1995) for most loci. The total PI value obtained for the complete set of STMS (2.06 x 10−12) provided confidence that plant samples showing the same allelic combinations in all the analyzed loci belonged to the same rootstock cultivar. For all loci, except ssrVrZAG79, the observed heterozygosity was lower than the expected heterozygosity.
The STMS analysis identified 92 genotypes among the 208 samples that supposedly represented 83 cultivars or rootstocks (Table 3⇓). Based on these results, five different types of results were observed that were grouped in five classes: (I) cultivars represented by accessions from a single source showing unique profile in the sample (e.g., Tomantjo V. riparia and Gaillard V. rupestris); (II) identical STMS genotypes between accessions of the same cultivars from different origin (e.g., 34 E.M. Foex, 99 Richter, and 1103 Paulsen), as expected from the clonal nature of the rootstocks cultivars; (III) identical STMS genotypes between accessions of different cultivars (e.g., 43DE of V. riparia Portalis rouge and 57IT of Gloire de Montpellier shared genotype 68); and (IV) different STMS genotypes for accessions of the same cultivar. These cases can be further differentiated into two subclasses: (IVa) when the unexpected genotype is unique (e.g., 140 Ruggeri, where genotype G38 is unique) or (IVb) is shared with accessions of another cultivar (e.g., the genotype G47 is shared between 216-3 Castel and G-1 Grézot). Samples included in class IV showed on average discrepancies for 10 alleles, ranging from the 17 differential alleles found in V. riparia var. Grand glabre (with genotypes G2 and G4) to the three differential alleles found in 8B Téléki (G30, G31, and G32) and ARG1 (G57 and G61) (Table 1⇑, Table 3⇓).
Since total paternity exclusion probability (Weir 1996) estimated for all STMS loci was very high (0.99999; Table 2⇑), compiled pedigree information (Galet 1956, Ravaz 1902) was used to try to identify the true genotype among samples in classes I, III, and IV and to confirm pedigree information of samples in class II. Complete pedigree and parental information was available for 16 of the hybrid accessions analyzed. Genetic and bibliographic information was in agreement for only two of these cases: 775 Paulsen (genotype G41) V. berlandieri Rèssèguier no2 (G13) x [V. rupestris du Lot (G7)] and 216-3 Castel (G46) [1616 Couderc (G24) x V. rupestris du Lot (G7)]. Additionally, for seven accessions the genotype of the sample and the genotype of one of the cited parents was consistent: 196-17 Castel (G45) with V. riparia Gloire (G68); Fercal (G79) with BC1 (G82); 5A Martínez-Zaporta (G80) with 41B Millardet et de Grasset (G14); 140 Ruggeri (G39) and 1103 Paulsen (G35) with V. rupestris du Lot (G7); and 3306 Couderc (G16) and 3309 Couderc G17 with V. rupestris Martin (G8).
Using this data we first tried to identify the true-to-type genotype of each accession. We next tried to identify the samples that did correspond to a true-to-type genotype and thus identify the accessions that likely had a false name. In considering the microsatellite profiles of plants maintained as the same cultivar or rootstock in different banks, the results of the pedigree analysis, the passport information, and the genotypic data published by other authors (Lamboy and Alpha et al. 1998, Lin and Walker 1998, Sefc et al. 1998, This et al. 2004, Thomas et al. 1994), it was possible to assign a specific STMS genotype to most of the analyzed accessions (Table 1⇑). When data from different sources did not agree, the most frequent genotype was chosen. Furthermore, genotypes with pedigrees that matched published information were considered to be true-to-type. Fifty-four percent of the samples were considered as true-to-type (113 plants, representing 46 accessions, indicated in bold in Table 1⇑).
For the remaining accessions, mainly included in class I, it was not possible to establish true-to-type genotype because the available material came from a single source, there was no published information on STMS genotypes, or STMS profile comparison revealed possible errors.
AFLP analysis.
Three AFLP primer combinations were tested on a subset of the samples. These combinations amplified 383 fragments, with sizes ranging from 35 to 700 bp. Of these fragments, 247 (64%) were polymorphic and easily scorable and were used for further analysis. All the studied plants could be differentiated.
A matrix of presence-absence of AFLP fragments was generated and the Dice coefficient was used to study genetic similarities (GS) between pair of accessions. A graphical representation of the UPGMA clustering based on the GS matrix was generated (Figure 1⇓). The high cophenetic correlation obtained between the GS matrix and the co-phenetic matrix (r = 0.806) showed a good fit of the cluster analysis. Average level of intraspecific GS detected within the American species (0.72 for V. riparia, 0.64 for V. rupestris, and 0.68 for V. berlandieri) was lower than in V. vinifera (GS = 0.75). On the other hand, GS values observed between species were lower, ranging from 0.44 between V. rupestris and V. vinifera to 0.66 between V. berlandieri and V. riparia. However, hybrid groups showed higher genetic similarity values (Table 1⇑). Thus, vinifera x rupestris hybrids were the most similar, with an average GS of 0.83. They were followed by hybrids derived from crosses between American species (berlandieri – rupestris, GS = 0.78; berlandieri – riparia, GS = 0.76; and riparia – rupestris, GS = 0.72).
Most accessions were grouped in two well-defined clusters (Figure 1⇑). The first included accessions of the American species (V. berlandieri, V rupestris, and V. riparia) and hybrids among them. The second cluster grouped all accessions of V. vinifera and the hybrid accessions derived from crosses between V. vinifera and any of the American species. The first cluster was divided in two subgroups. One included 20 accessions: four cultivars of V. berlandieri, 15 hybrids derived from crosses between V. berlandieri and other species, and accession 14IT. This accession corresponds to the hybrid 101-14 Millardet, which has been described as a cross between unknown cultivars of V. riparia and V. rupestris (Galet 1956). The second subgroup of the first cluster included 24 accessions: nine cultivars of V. riparia, seven of V. rupestris, seven hybrids derived from crosses involving one or both species, and accession 30FG (31 Richter, described as a hybrid Rèssèrguier no2 of V. berlandieri x Novo Mexicana). On the other hand, the second cluster grouped the seven V. vinifera cultivars analyzed; the single variety of V. labrusca analyzed; and the 10 hybrids derived from crosses between V. vinifera and American species (V. rupestris, V. berlandieri, V. riparia). However, it also grouped the cultivars Vergisso of V. berlandieri, Ganzin of V. rupestris, and two American rootstocks: 228-1 Castel (17ESS1), derived from the cross Solonis x du Lot (V.rupestris), and 196-17 Castel (31ESS1), a hybrid from the cross Gloire (V. riparia) x 1203 Couderc. Accession 71ESS1, corresponding to hybrid 5A Martinez-Zaporta, documented as an autopollination of 41B Millardet Grasset (V. berlandieri x V. vinifera), remained ungrouped.
Discussion
Grape rootstocks are a numerous and heterogeneous group of plants used to graft cultivars of V. vinifera. They consist of different cultivars of American Vitis species (commonly V. rupestris, V. riparia, and V. berlandieri) and hybrids between them (American-American hybrids) and with V. vinifera (American-European hybrids). A few hybrids also derive from complex crosses among three or more species (Galet 1956). As a first approach to analyze the extent of genetic diversity of rootstocks, STMS and AFLP markers were used to study accessions of main rootstock cultivars. The use of nine STMS developed for V. vinifera enabled genotyping of all tested samples, demonstrating their usefulness in the identification of Vitis species and hybrids. Moreover, all the genotypes identified could be differentiated using a minimal set of four markers: ssrVrZAG47, VVMD7, VVS2, and VVS1. Loci ssrVrZAG47, VVMD7, and VVS2 were the most informative, with an effective number of alleles (Ne) of 13.5, 10.6, and 10.3, respectively. VVS1 is a highly heterozygous locus, but the existence of alleles differing in only a base pair hampers the analysis.
When all samples were considered, the heterozygosity obtained with STMS was 0.78. This value is higher than that obtained for V. vinifera cultivars (Ibañez et al. 2003), but it is in the same range as that obtained when a set of rootstocks is included (Lin and Walter 1998, This et al. 2004). In a study of a sample of 46 cultivars (including 15 rootstock cultivars) selected to maximize the allelic diversity, 107 alleles were identified for the six analyzed loci (VVMD5, VVMD7, VVMD27, VVS2, ssrVrZAG62, and ssrVrZAG79), with an average of 17.8 alleles per locus (This et al. 2004). Similar values were detected with the set of STMS markers tested here (105 alleles, with an average of 17.5 per locus) for the same six loci, since locus ssrVrZAG47 is the same as VVMD27 (Pollefeys et al. 2003). These numbers include two new alleles detected for locus VVMD5 (1MG1+2 and 11R2+2), four for ssrVrZAG62 (CF2+2, CF2+4, 11R2+2, and 11R+4), one for VVMD27/ ssrVrZAG47 (GO2+2), and one for ZAG79 (99R2 +2).
Comparison of genotypes obtained for different plants of the same cultivar or hybrid rootstock allowed the identification of misclassified samples. The detection of different genotypes in plants belonging to a single cultivar is probably due to mistakes in the classification or maintenance of these materials. The identical STMS genotypes obtained for accessions of different cultivars, however, could be explained by the presence of synonymies, such as V. rupestris Saint George/V. rupestris du Lot, or duplications among the analyzed samples. Moreover, cultivars represented by accessions from a single source could also include some mistakes. In any case, the genotypes assigned to the accessions do not show important discrepancies with published data (only one allele in one accession; Table 1⇑), with the exception of Lin and Walker (1998). Out of the 29 cultivars in common with these authors, 11 showed differences (Table 1⇑). This difference could be related to the existence of different genotypes with the same accession name in different germplasm banks. For example, our work analyzed four accessions of 1616 Couderc from different sources (16ES1/2, 16DE, 16IT, and 16FR) resulting in four genotypes. Whereas the 1616 Couderc genotyped by Lin and Walker (1998) matched with 16IT, and those tested by Sefc et al. (1998) and This et al. (2004) matched with 16FR. On the other hand, the genotypes obtained for the nine samples provided by a private breeding group in order to verify commercial rootstocks cultivars matched with their respective true-to-type cultivars (Table 1⇑).
These results point out mislabeling, multiple naming of a single cultivar, and incorrect ampelographic identifications in rootstock collections. The number of mislabeled accessions suggests that these mistakes have been produced during many years of vegetative propagation. For example, at least three different forms of rootstock 1616 Couderc were reported as early as 1947 (Manaresi 1947). In addition, there is contradictory information for some root-stock pedigrees. For example, three different origins have been suggested for Solonis: it is considered a hybrid (Galet 1956) or a variety of different Vitis species depending on the authors: V. solonis (Larrea 1978), V. longii (Foex 1888), and V. acerifolia (Moore 1991). Therefore, it is not surprising that only two of the 16 pedigrees studied agree with their proposed parentage: 775 Paulsen (V. berlandieri Rèssèguier no2 x V. rupestris du Lot) and 216-3 Castel (1616 Couderc x V. rupestris du Lot). Three different origins had been suggested for 216-3 Castel: 1616 Couderc x V. rupestris Lot (Galet 1956); 1616 Couderc x Solonis (Larrea 1978); and Solonis x Lot V. rupestris (Ravaz 1902). STMS analysis confirms that suggested by Galet. The low number of cultivars analyzed and the use of putative misclassified material, not only in this study but also for development of rootstock hybrids, could have led to a low success of pedigree discrimination with microsatellites.
The occurrence of occasional mistakes is unavoidable, especially with ancient clonal crops, and it has been observed that 5 to 10% of grape cultivars maintained in grape collections is not correctly annotated (Dettweiiler-Münch 1992). Although the number of accessions analyzed in our study was not high enough to consider the results as representative of the current state of rootstock germplasm collections, ~27% of accessions analyzed had problems in their classification. This percentage is likely higher, since some cultivars represented by accessions from a single source could also include mistakes. These mistakes are difficult to detect based on ampelographic descriptors and they reduce the practical value of collections, since they could lead to both duplication of accessions and loss of genotypes, leading to genetic erosion.
The use of three AFLP primer combinations provided a general view of the genetic relationships among the studied material, allowing the discrimination of all the studied rootstocks. Although the number of the accessions studied is not high enough to produce a good representation of the intraspecific genetic diversity, in general all cultivars belonging to same species grouped together. American species were clearly differentiated from European V. vinifera. Among American species, V. riparia and V. rupestris showed higher values of genetic similarity than with V. berlandieri, which grouped in a single cluster separated from the other accessions. These results agree with the classification proposed by Galet (1967), which cluster both species, V. riparia and V. rupestris, in the same series of Euvitis (series 10, Ripariae). This study does not provide sufficiently accurate estimates of intraspecific genetic diversity because only a few plants of each species have been studied. However, moderate-to-high genetic similarities were observed among accessions within species. Similar levels were found in a study of V. vinifera cultivars with AFLP (Cervera et al. 1998). The accession of V. labrusca showed high GS with V. vinifera cultivars. This result could be related with a possible hybrid origin of the studied accession. However, similar results were found elsewhere, where higher similarity was observed for V. labrusca with V. vinifera than with V. berlandieri, V. riparia and V. rupestris species, when studying the degree of conservation of 11 microsatellite sequences across Vitis (Di Gaspero et al. 2000).
When the hybrids are considered, the lowest GS AFLP values were found between riparia - vinifera hybrids, with GS values ranging from 0.62 to 0.74. These accessions (143 b2 Millardet, 141 A Millardet, and 595 Oberlin) appeared in separate clusters in Figure 1⇑. The other hybrid groups showed higher GS values (between 0.66 and 0.92) and hybrids with similar origin appeared close in the dendrogram. The group showing the widest AFLP GS range was berlandieri - riparia hybrids, which included the most similar cultivars based on AFLPs: 5 BB Téléki Sélection Kober and SO4 Téléki.
Although the clustering was consistent with the genetic origin of most of these cultivars, the position of some accessions was unexpected. The samples analyzed as representatives of cultivars Vergisso of V. berlandieri (42ESS1) and Ganzin of V. rupestris (61BG) presented higher levels of similarity with V. vinifera or V. labrusca than with their own putative specie. These results suggest that either both accessions are distantly related to other analyzed samples of corresponding species or they do not represent the true-to-type version of the accession. In the same way, variety Tomantjo of V. riparia (45B), which was not included in the V. rupestris - V. riparia cluster, showed lower GS values with the rest of V. riparia than with V. rupestris. The possible hybrid origin of several cultivars has been suggested. For example, some cultivars of V. rupestris have been described as putative hybrids (Metallique as a hybrid of V. rupestris x V. candicans, and Galliard and du Lot as hybrids with other Vitis species, such as V. monticola or V. cordifolia) (Ravaz 1902). Other accessions that clustered in unexpected positions were 31 Richter, 101-14 Millardet, 228-1 Castel, 196-17 Castel, and 5A Martínez Zaporta.
Conclusion
Microsatellite genotypes have been identified for a large sample of rootstocks, and misclassification problems have been detected that genetic resource centers will have to address. This work also demonstrates how a combined approach based on the study of AFLP-based genetic relationships together with microsatellite genotyping and pedigree analyses is required to unravel the origin and the trueness-to-type of most of the current rootstock accessions. Such efforts are required to make these genetic resources fully useful.
Footnotes
Acknowledgments: The authors thank the different Vitis germplasm banks used in the study, and the private breeding group PROVEDO S.A. for providing the analyzed accessions.
M.T.A and J.A.C. were funded by fellowships from INIA. This research was supported by INIA Project RF99.009.
- Received August 2005.
- Revision received February 2006.
- Revision received July 2006.
- Copyright © 2007 by the American Society for Enology and Viticulture