Introduction

The genus Vitis L. is diverse, comprising 40–60 species in Asia, about 25 in North America and a single European species—Vitis vinifera L. This last one is the main species cultivated today, while the other Vitis species are used mostly for breeding rootstocks and fungus-resistant scion cultivars. In V. vinifera L., around 6,000 cultivars are estimated to exist (Alleweldt and Dettweiler 1994), of which less than 400 are of commercial importance (Galet 2000). Therefore, today most of the genetic resources of V. vinifera L. are maintained in germplasm collections.

Grapevines have been cultivated for about 5,000 years (Zohary and Hopf 2000). The ease of vegetative propagation has favored widespread diffusion of many cultivars to diverse regions of the world (Dion 1977; Fregoni 1991). As a consequence, some cultivars now have up to 100 synonyms, and numerous homonyms also exist (http://www.genres.de/idb/vitis/). Because accurate identification of accessions is a basic requirement for the rational management and use of germplasm, the clarification of synonymy, homonymy, and misnaming is a significant problem in the 130 grapevine collections that exist worldwide (Dettweiler et al. 2000a).

The identification of grape cultivars has traditionally been based on ampelography (from the Greek ampelos—grapevine and graphos—description), which is the analysis and comparison of morphological characters of leaves, shoot tips, fruit clusters, and berries (Galet 1991; Boursiquot and This 1996; IPGRI UPOV OIV 1997; Galet 2000). Expertise in ampelography, however, is restricted to a small and declining number of specialists. Additionally, the expression of morphological characters is influenced by environmental factors, individual plant biology, and life history. Furthermore, juvenile plants are nearly impossible to identify because within 4 or 5 years, they do not exhibit the typical morphological traits of adult plants. Some genetically related cultivars are morphologically very similar and difficult to differentiate by visual comparison (Aradhya et al. 2003). On the other hand, intravarietal clones can differ considerably in phenotype even though they have virtually identical DNA profiles (Vignani et al. 1996; Franks et al. 2002; Riaz et al. 2002).

To surmount these limitations, molecular markers have been used to differentiate, characterize, and identify grapevine accessions. RFLP (Striem et al. 1990; Bourquin et al. 1993; Bowers and Meredith 1996), RAPD (Grando et al. 1995; Loureiro et al. 1998; Ye et al. 1998; Tessier et al. 1999), AFLP (Sensi et al. 1997; Cervera et al. 1998), and microsatellite markers (Botta et al. 1995; Lin and Walker 1998; Sefc et al. 2000; Aradhya et al. 2003) have all proven useful. Microsatellite markers are favored, however, because of their combination of polymorphism, reproducibility, and their codominant nature (Sefc et al. 2001).

GENRES081 was a European Union research project focused on the compilation, standardization, and exchange of information concerning grapevine genetic resources (Dettweiler et al. 2000b; This and Dettweiler 2003; http://www.genres.de/vitis/). The partners in the project, representing the major European grapevine collections, set as an objective the development of a central European database containing reference microsatellite profiles for true-to-type identification of grapevine accessions. However, as demonstrated in tomato (Bredemeijer et al. 2002) and wheat (Röder et al. 2002), for such a database to be useful to diverse laboratories employing differing equipment and methods, alleles must be standardized. Thus, the main objectives of the present study were: (1) to compare different methods of microsatellite DNA profiling for reproducibility among the partners and (2) to standardize allele scoring by defining reference alleles. Six informative markers were selected, and all participants analyzed identical DNA samples. Because of the diversity of laboratory equipment and protocols, no standardization of PCR protocols was attempted. A set of coded alleles based on well-known reference cultivars was developed that facilitates data comparison among laboratories and will permit the development of a common international database.

Materials and methods

Cultivars

A total of 47 grapevine accessions were analyzed, corresponding to 46 different cultivars (‘Merlot’ was represented twice since material from this cultivars originated from two different germplasm collections, Table 1). For the names of the cultivars, “N,” “B,” and “RG” refer to berry skin color—black, white and red, respectively. The number and choice of cultivars analyzed reflect the evolution of the microsatellite project within GENRES081 over the course of three international workshops held in 1998, 1999, and 2001 (http://www.genres.de/vitis).

Table 1 Grapevine cultivar accessions analyzed in this study. Trueness-to-type was defined by ampelography
Full size table

The first round of analyses was limited to five regionally important cultivars in order to obtain preliminary information on data consistency (Table 1; analysis group 1). These cultivars were expected to be quite distinct in allelic profile because of their different geographic origins.

In an effort to reach better agreement, in a second round, ten well-known cultivars were analyzed (Table 1; analysis group 2). Cultivars from the first round were also analyzed in round 2.

In a third round designed to complete the allelic ladders for the six microsatellite loci, 34 additional cultivars, including 15 rootstock cultivars, were selected (Table 1; analysis group 3). This group was chosen after consulting three existing grapevine microsatellite databases in order to maximize allele diversity (Sefc et al. 2000; E. Zyprian and C. Meredith, unpublished data). Whenever possible, well-known and widely distributed cultivars were chosen in order to maximize the value of this work to the scientific community.

Plant material

To prevent confusion concerning trueness-to-type of the analyzed accessions (Dettweiler et al. 2000a), it was important to insure a single source for each sample in the analysis, particularly in the second and third round. Trueness-to-type was defined by comparing morphological descriptions and photographs to the documentation of morphological characters in the literature. For each accession used, fresh young leaves or late-winter cuttings of true-to-type vines were collected from only one germplasm collection (Table 1). For the first and second rounds, DNA was extracted from leaves by single partners and distributed to all the others. For the third round, cuttings were collected from the Institut National de la Recherche Agronomique Vassal domain and sent to each partner for local DNA extraction. DNA from these cuttings was sent to the Meredith laboratory in the United States because of quarantine restrictions.

Marker selection

In 1998, VVS, VVMD, and VrZAG markers were the most widely used grapevine microsatellite markers (Cipriani et al. 1994; Thomas et al. 1994; Botta et al. 1995; Bowers et al. 1996; Regner et al. 1996). Based on the experience of the project partners, six polymorphic markers were chosen: VVS2 (Thomas and Scott 1993), VVMD5, VVMD7 (Bowers et al. 1996), VVMD27 (Bowers et al. 1999b), VrZAG62, and VrZAG79 (Sefc et al. 1999).

DNA extraction

DNA was isolated from fine powdered leaf or cambium tissues frozen in liquid nitrogen and ground in a mortar. Partners (partner no. according to title page) used a number of different DNA extraction protocols, including those of Doyle and Doyle (1990) with an additional RNAse A-digestion step (partner 5), Doyle and Doyle (1990) modified by Cipriani et al. (1994) (partner 10), Thomas et al. (1993) (partner 3), Thomas et al. (1993) without the initial step (partner 8), Crespan et al. (1999) (partner 6), Ferreira Monteiro et al. (2000) (partner 9), or according to the protocol for DNeasy Plant Mini Kit (Qiagen, Hilden, Germany) (partner 2, partner 4), or the latter with addition of 1% Polyvinylpyrrolidone 40 (Sigma-Aldrich, Dearborn, Mich., USA) to AP1-buffer and doubling the amount of AE-buffer for elution (partner 1). Specific conditions are available on request from this@ensam.inra.fr.

PCR conditions

According to each partner’s preferences for different Taq DNA polymerases and thermal cyclers, various individual strategies for optimization and generalization of PCR conditions were employed (specific conditions are available on request from this@ensam.inra.fr). PCR reaction mixes and cycling strategies differed widely, but there was a tendency towards hot-start PCR. Most partners preferred a three-step cycling routine, but partners 6, 7, and 8 used differing two-step PCR regimes. Partner 10 performed a touchdown PCR, differentiated as two separate protocols: (1) for VVS2, VrZAG62, VrZAG79, touchdown PCR of −1°C/cycle from 65 to 56°C and (2) for VVMD5, VVMD7, VVMD27, touchdown PCR of −1°C/cycle from 65 to 52°C. Some partners generalized primer-specific annealing temperatures to reduce the number of cycling programs.

Microsatellite analysis

Different systems for fragment differentiation and allele size determination were also employed. Six partners separated amplification products manually by high-voltage electrophoresis on vertical, denaturing polyacrylamide sequencing gels, and four partners utilized single capillary or gel-based electrophoresis systems for automatic fragment separation in ABI Prism Genetic Analyzers (Applied Biosystems, Foster City, Calif., USA). Precise conditions are available on request from this@ensam.inra.fr.

Manual analysis

Polyacrylamide gels (5–6% polyacrylamide, 7–8 M urea) were prepared according to standard protocols (Sambrook et al. 1989). Before loading, samples were denatured for 2–3 min at 94–95°C in solutions containing balanced amounts of PCR amplification products in a buffer of formamide or FBX-marker (100 μl 1 M NaOH, 400 μl H2O, 9.5 ml formamide, 50 mg Bromophenol Blue, 50 mg xylene cyanol). After electrophoresis, bands were visualized by a silver-staining procedure, according to the protocol provided with the Silver Sequence DNA Sequencing System (Promega, Madison, Wis., USA). Partner 6 performed silver staining according to Crespan and Milani (2001), following Bassam et al. (1991) and Tixier et al. (1997), with the exception of using NaOH instead of NaCO3. Partner 10 used γ[33P]-ATP-labeled primers and detection of PCR fragments on autoradiographic film after 1–7 days of exposure to fragment radiation.

Automatic analysis

For automatic electrophoresis (single capillary on ABI 310 or gels on ABI 373 and 377), the amplification product was added to 12–20 μl of deionized formamide. Adapted amounts of denatured PCR fragments labeled with fluorescent dye phosphoramidites were loaded, separated on the capillaries or the gels, and detected by the system’s laser. Fluorescent emission was analyzed by GeneScan software, version 2.1 (Applied Biosystems), using internal-lane size standards (ROX or TAMRA) and the system’s local southern method for automatic size calling of peak positions.

Fragment-size determination and data analysis

Fragment-size determination for manual analysis was adjusted by the use of PCR fragments of previously analyzed cultivars as internal size standards, in addition to commercial weight markers. Once the fragment sizes of reference alleles had been determined, partners 2 and 6 used exclusively self-made primer-specific allelic ladders.

The automatic peak labeling of the ABI Prism Genetic Analyzers needed some additional control by visual inspection of individual peak positions, rounding up or down the decimal places to reach integer allele numbers. Previously labeled peak profiles of reference cultivars aided decision making. Partner 4 applied the algorithm of Ghosh et al. (1997), using the common average value of identical fragments to direct the algorithmic rounding of decimal variations into same integer numbers except for VVMD27, which was rounded manually because some of its alleles differ only by 1 bp.

Polymorphism information content (PIC) and discrimination power were calculated according to Lamboy and Alpha (1998).

Results and discussion

Although a considerable amount of microsatellite data has been published for grape, interlaboratory variations have made comparisons of results difficult, more particularly because grapevine is highly heterozygous. In this study, we have compared microsatellite data for 46 different cultivars with six microsatellite markers among ten different laboratories and implemented a reference allele system for uniform grapevine microsatellite analysis.

All data from this work, including detailed data from each partner are available upon request to this@ensam.inra.fr or at http://www.genres.de/eccdb/vitis/

Comparison of allele sizes

Comparison of the allele sizes obtained by the different laboratories for each DNA sample for the six microsatellite markers produced no satisfactory agreement (Table 2) as it has been previously reported for other crops (Haberl and Tautz 1999). Although some partners obtained similar results, in other cases the raw data for identical alleles differed by as much as 5 bp (e.g., VVMD5 or VVS2, Table 2). The magnitude of the difference varied by marker (e.g., no more than 3-bp differences for VVMD7), but in general, the raw data (without any transformation) could not be compared. In order to increase the general applicability of this work, we had previously decided not to standardize the protocols. Each of the ten partners analyzed the same DNA with its own laboratory equipment and individually adapted protocols, so some of the differences may have been the consequence of the different protocols. No clear relationship could be found, however, between any protocol step and the fragment size. Automatic sequencing does tend to give smaller fragment sizes, but comparable sizes were obtained with both automated and manual techniques. Since direct comparison of profiles was impossible without an additional harmonization procedure, we tested several methods, for comparing data between laboratories.

Table 2 Examples of profiles obtained by the ten partners for two cultivars displaying identical coded data among the partners. Allele sizes are in base pairs
Full size table

Transforming numerical data

The relative differences between the two fragments of an allele pair were first considered. The data were coded as size differences between the smallest allele observed in the sample (n) and size of the allele of the cultivar. For example for partner 1, ‘Mourvèdre’ was 134/152 for VVS2, and was coded n+10/n+28 (Table 3) since the smallest allele, n, was 124. With this method, results from the different laboratories were more consistent. But discrepancies still could not be avoided for all the markers, and shifts in the relative difference between the alleles occurred (Table 3). The differences were mostly the result of the rounding methods. The simple mathematical algorithms employed in the automatic scoring of peak sizes can produce artificial shifts by automatically rounding up. Furthermore, this strategy depends on the size and composition of the sample: according to the sample analyzed by different laboratories, the smallest allele observed (n) might vary, and comparison of data between laboratories would be difficult. This strongly argues against this strategy. Thus it was decided to concentrate on the development of a standardized coding procedure. For each SSR marker, the profiles of the 46 cultivars were compared and PCR fragments (alleles) were sorted according to their lengths. We thus selected one PCR fragment of each size represented in order to cover the entire size range of detected fragments. These selected fragments were designated as reference alleles and assigned code names based on the cultivar in which they were observed. If the fragment were the shorter or the longer for this cultivar, it was designated “1” or “2,” respectively. For example, the shorter fragment of ‘Cabernet-Sauvignon’ was assigned the code CS1. The cultivars selected as references, including both scion and rootstock cultivars, are presented with their codes in Table 1. For most of the reference alleles, there was a consensus between the partners. For VVMD27, the data were identical. In very few cases (2.4%) some discrepancies arose (i.e., one partner found one or several additional reference alleles), particularly for VrZAG62, VrZAG79, and VVMD7, but we discarded them as genotyping errors. The use of these cultivar-specific fragments as size standards produced a homogenous coding system, comprising a relatively complete allelic ladder for each of the six microsatellite loci (Table 4).

Table 3 Comparison of profiles at locus VVS2 after codification, using relative size differences between smallest allele detected in the sample (n) and size of the allele of the cultivar
Full size table
Table 4 List of the reference alleles for each of the six loci, with a general indication of their size. The size is also given as relative size to n. The codes are as indicated in Table 1. Numbers after the codes refers to shortest (1) or longest (2) allele of the reference cultivar. The numbers are given as indication and correspond to the most common size among the partners. Rootstocks alleles are shown in italics
Full size table

The number of reference alleles among the 33 reference cultivars ranged from 13 for VrZAG79 to 23 for VVMD27 (Table 4). The relative difference in size between the shortest and the longest alleles ranged from 26 bp for VrZAG79 to 46 bp for VVMD5 and VrZAG62. For VVMD7, VVMD27, VVS2, and VrZAG79, the allelic ladders were almost complete, with nearly every expected size increment observed, whereas more gaps were observed for VVMD5 and VrZAG62, notably for the larger allele sizes. A few of these missing alleles were observed, but synonymy of cultivar names could not be resolved. By consequence, no true-to-type accession could be proposed as a reference.

In most cases, allele size increments were 2 bp, in accordance with the dinucleotide nature of these markers (Thomas and Scott 1993; Bowers et al. 1996, 1999b; Sefc et al. 1999). For VVMD27 and VrZAG62, however, a few 1-bp increments were repeatedly recorded (n+11 and n+19 for VVMD27 and n+1 for VrZAG62). Such fragment patterns might be interpreted as stutter or as the extra base additions that occur with some Taq polymerases (Brownstein et al. 1996), or they might even be completely ignored. In our case, however, comparison of data over laboratories confirmed these differences.

Special attention was given to the selection of the reference cultivars. Preference was given to well-known cultivars, except for a few alleles that were detected only in rare cultivars. The total number of cultivars in the reference set was minimized by selecting cultivars useful for more than one locus. In order to fulfill these criteria and to cover the complete range of alleles, ten well-known cultivars [‘Barbera’, ‘Cabernet-Sauvignon’, ‘Cabernet franc’, ‘Chardonnay’, ‘Merlot’, ‘Muscat à petits grains blancs’, ‘Pinot noir’, ‘Sultanina’ (or ‘Thompson seedless’), ‘Silvaner’, and ‘Traminer rot’ (or ‘Gewürztraminer’’)] were initially selected. Seven well-known rootstocks (‘Couderc 1616’, ‘Couderc 3309’, ‘Millardet et Grasset 101-14’, ‘Millardet et Grasset 420A’, ‘Richter 99’, ‘Richter 110’, and ‘Teleki 5C’) were also added to the set. The inclusion of rootstock cultivars into the analysis was necessary, not only because of their general importance in grapevine cultivation, but also because of the high number of exclusive alleles not found among V. vinifera cultivars. The rootstocks frequently displayed characteristic allele clusters that were either smaller (e.g., VrZAG62) or larger (e.g., VVMD5 and VVMD27) compared to those that were detected in V. vinifera cultivars. In order to fill most of the remaining gaps, a few less well-known scion cultivars were added too. The 33 reference cultivars are available upon request to didier.vares@ensam.inra.fr.

The extension of the allele ranges for these markers will also be of great importance for the analysis of other Vitis species and interspecific hybrids. For example, in a work on 105 accessions of 16 American Vitis species (found in the pedigrees of almost every known hybrid), 25 and 23 alleles were observed for VVS2 and VVMD7, respectively (Lamboy and Alpha 1998). The allele ranges, however, were very similar compared to our work. Very few rare alleles (three for VVMD7, five for VVS2), present in a few species only (V. Vulpina L., V. palmata Vahl, V. Piasezkii Maxim., V. arizonica Engelm.), were outside the size range represented in our reference set.

Comparison of coded data

For each cultivar and each marker, the data were coded according to the defined reference alleles and compared. Complete data are available on request to this@ensam.inra.fr or at http://www.genres.de/eccdb/vitis/.

Compared to the raw data (Table 2) and relative allele size differences (Table 3), coding of the numerical data allowed the immediate and direct comparison of data. In fact, coded data were easy to compare and, except for a few discrepancies, were identical for all partners. For four cultivars (‘Hans RG’, ‘Malegue 44–53’, ‘Mancin N’, and ‘Mourvèdre N’), the data were completely identical. For the others, in general only one partner showed discrepancies for one or two loci. The “true” allele was thus deduced from the data that were identical among most of the partners. The discrepancies were then calculated from these “true” profiles (Table 5). Excluding missing data, 97.5% of the alleles (4,487 alleles out of 4,600) were completely identical among partners (Table 6). Data homogeneity was especially high for VVS2, VVMD27, and VVMD7 (98.8, 98.6, and 98.0%, respectively). On the other hand, the results for VrZAG62 and VrZAG79 were in agreement for 96.4% and 96.2% of the alleles, respectively.

Table 5 Summarized coded microsatellite profiles for 46 grapevine cultivars uniformly coded by the PCR fragments of selected reference cultivars
Full size table
Table 6 Consistency of the data over the partners. Each allele for the 44 cultivars analyzed in round two and three was compared between each partners
Full size table

VVMD5, VVMD7, and VVMD27 are robust markers with stable, clear fragment patterns. VVS2 was easy to score since all alleles are represented in the reference set. On the other hand, VrZAG62 and VrZAG79 could produce interpretation problems because of their tendency to produce unclear banding patterns or stutter. Three types of discrepancies were observed. In a few cases, the data were very different. These discrepancies most probably correspond to typing errors. Most differences were consistent 1- or 2-bp shifts for several alleles, either in the standards or in the analyzed cultivars. Finally, some discrepancies occurred in the case of single alleles that are only 2 bp apart.

Data discrepancies due to shifts are difficult to explain. Many were avoided by covering almost the complete spectrum of alleles detected for each of the loci. In any case, they showed the necessity to optimize PCR conditions and visualization techniques in order to avoid conditions that can lead to stutter or fluctuating fragment patterns (Hu 1993; Smith et al. 1995). A good strategy is the general use of a hot-start Taq, the optimization of annealing temperatures for each primer pair, and the use of cultivar-specific, coded reference fragments as universal size standards. Using pigtail primers (Bredemeijer et al. 2002; Röder et al. 2002) can also help to circumvent these problems.

Discrepancies in single alleles only 2 bp apart are easier to explain than to avoid. They are likely due to misinterpretation of homozygous versus heterozygous state of cultivars when the alleles are only 2 bp apart (one microsatellite repeat unit), and the microsatellite locus also shows either extra base additions or stutters. Heterozygosity in grape is very high and extra difficulties have been described when working with heterozygous plants (Vosman et al. 2001). No easy way to reduce this type of error can be proposed. Automatic sequencing seems less subject to these errors but they arise nevertheless.

The percentage of consistency reported here is similar to that reported for tomato (Bredemeijer et al. 2002). In comparing data between only two laboratories, the authors reported 97.3% concordance, slightly less than the 99.5% reported for wheat (Röder et al. 2002) with generalized PCR conditions.

Since the illustrated method enabled easy comparisons among the partners, we propose that other labs utilizing grapevine SSR fingerprints convert to the code. This could be achieved by identifying the size of the reference alleles in their database, and converting all data sharing the same size, using the proposed codes (Table 4). An access database was developed in order to help this conversion and is available on request to this@ensam.inra.fr.

Development of the database

The data for the 46 analyzed cultivars were verified by seven to ten partners, so the database developed in this study is very strong. It also includes several major cultivars that are grown worldwide and will be highly useful as multiconfirmed reference for identification purposes.

The six selected microsatellite markers are suitable for grapevine cultivar characterization because of their high degree of allelic polymorphism (PIC varies from 0.86 for VVMD7 to 0.91 for VVMD5 in our study) and high discrimination power (ranging from 0.95 for VrZAG62 to 1 for VVS2). Because they are already in wide use, these six markers should be recommended generally as the minimal standard marker set for future grapevine-cultivar analyses. That would facilitate the creation of uniform, easily comparable data catalogs for the comparative identification of unknown or unconfirmed accessions in international grapevine germplasm collections. In all cases, the six markers turned out to be sufficient to differentiate each of the 46 cultivars in this study by a clear, individual allelic profile. As few as two markers (e.g., VVMD27 and VVMD5) were sufficient for the discrimination of the cultivars, but more loci were chosen in order to increase the polymorphism and thus reduce the probability of false identity. Since this project began, the number of available grape microsatellite markers has rapidly increased to about 400. Not all those markers are necessary for identification, and there is no question that not all these markers can be standardized by a comparative international research effort. But for the purpose of general data comparability across laboratories, every scientist working with additional microsatellite markers would be advised to select only clearly expressed and highly polymorphic markers and to define a set of cultivar-based reference alleles that represent the allelic ladder as completely as is practical. This would enable others to easily convert the data into comparable data catalogs.

Conclusion

The objective of the study was to test the general comparability and reproducibility of microsatellite data produced by different laboratories under varying local conditions. If the characterization of grapevine cultivars by microsatellite fragment analysis works universally and independently from analysis systems and laboratory equipment, each analysis of identical DNA samples must produce identical allelic profiles.

In the present study, our only requirement was the analysis of identical samples, in order to avoid problems of sample identity as observed by Röder et al. (2002) and Bredemeijer et al. (2002). DNA from 47 samples was then analyzed with six microsatellite markers. To allow comparability of the data among the partners, the data were coded according to reference alleles from 33 cultivars. To reduce discrepancies due to stutter and fluctuating fragment patterns, the optimization of protocols individually adapted to local laboratory equipment must be completed for each individual marker.

The most frequent alleles for all six markers can be provided by a set of 17 cultivars as follows: ‘Barbera’, ‘Cabernet-Sauvignon’, ‘Cabernet franc’, ‘Chardonnay’, ‘Merlot’, ‘Muscat à petits grains blancs’, ‘Pinot noir’, ‘Sultanina’, ‘Silvaner’, ‘Traminer rot’, ‘Couderc 1616’, ‘Couderc 3309’, ‘Millardet et Grasset 101-14’, ‘Millardet et Grasset 420A’, ‘Richter 99’, ‘Richter 110’, and ‘Teleki 5C’. For applied viticulture studies, fundamental grapevine biology research and for management of grapevine germplasm, the accurate identification of cultivars is essential. The establishment and feeding of a uniform database with confirmed microsatellite profiles for true-to-type grapevine cultivar would support better and more rationalized management of grapevine collections.

Although the 46 cultivars employed in this study represent a small proportion of total cultivars, they nevertheless were chosen in order to be highly diverse. A total of 107 alleles were detected with the six markers (from 13 to 23 alleles each) with a mean value of 17.8 alleles per locus. This value is very high compared to other data published on grape. Bowers et al. (1999a) detected a mean of 11 alleles per locus with 350 French cultivars, and Sefc et al. (2000) detected a mean of 9.8 alleles per locus with a set of 164 cultivars. This number is, however, in the same range as was observed with a set of 58 rootstocks and one V. vinifera cultivar by Lin and Walker (1998) since they detected a mean of 17.6 alleles per locus. Our results were achieved by combining V. vinifera L. cultivars with interspecific hybrid rootstock cultivars.

The cultivars analyzed in this study include some of the world’s major wine cultivars (e.g., ‘Cabernet-Sauvignon’, ‘Chardonnay’, ‘Merlot’), an internationally important table grape and raisin cultivar (‘Sultanina’), and some of the most important and widely grown rootstocks. Thus, the results presented here are likely to be of value in all grape-growing regions of the world.

We have used in this work only six loci, not all of which are localized on different linkage groups (Riaz et al. 2003). Complete and irrefutable identification will often require a larger number of loci. However, the main goal of this work was to facilitate the comparison of grape microsatellite data among laboratories. In doing so, we have thus established the foundation for an international grape-cultivar database that will benefit the entire grape community.

We have demonstrated the usefulness of the coding procedure. We encourage other communities working on genetic resources to develop a similar method of defining common and unique reference allele set (representing a large range of alleles and a ladder as complete as possible) for several SSR loci. This set of reference alleles can then be used in order to code the data from several laboratories and would enable a very easy comparison of data and/or genetic resources.