Genetic Structure and Relationships among Wild and Cultivated Grapevines from Central Europe and Part of the Western Balkan Peninsula

The genetic diversity and relationship between wild (Vitis vinifera L. subsp. sylvestris (Gmel.) Hegi and cultivated (V. vinifera L. subsp. vinifera) grapevine in the western Balkan region and Central Europe have not been studied together previously, although this area has a rich viticultural past. Here, we studied wild grapevine populations sampled from their natural habitats in several countries of the western Balkan region and Central Europe. Their genetic diversity and structure were compared to cultivars that are traditionally in use in this region. A sample set of 243 accessions was genotyped at 20 nuclear microsatellite loci, including 167 sylvestris and 76 diverse vinifera cultivars. The genetic diversity of the wild grapevines was lower than that of cultivars by all genetic parameters. Both hierarchical and nonhierarchical clustering methods differentiated two main groups, indicating clear separation between wild and cultivated vines but also revealed clear gene flow between the cultivated and wild gene pools through overlaps and admixed ancestry values in the graphs. There was greater affinity to the wild grapes in Central European cultivars than in Balkan cultivars. Fine arrangement of the structure among cultivated grapevines showed differentiation among Central European and Balkan cultivars. These results confirm the divergence of wild grapes from vinifera and highlight the “crossroad” role of the western Balkan peninsula in the broader context of European viticulture.


Introduction
Wild grape (V. v. subsp. sylvestris) is a close relative of the cultivated grapevine (V. v. subsp. vinifera) and the only endemic taxon within the Vitaceae family in Europe [1]. Research on wild grape intensified at the begining of the 21st century concurrenty with improvement in molecular identification methods, with the primary goals of conserving its biodiversity, clarifying its taxonomic status and identifying traits of interest for grapevine breeding. Today, populations of V. sylvestris are endangered and fragmented, in some cases down to <10 individuals [2] or almost completely devastated [3][4][5]. The natural habitats and populations of wild grapes stretch from the Atlantic coast to the western Himalayas and from the Rhine valley to Tunisia [6,7], coexisting with cultivated varieties [1]. In some European countries, V. sylvestris is an officially protected species (Spain, France, Germany, and Hungary), while in Croatia, it has "least critical (LC)" status according to the Flora Croatica Database [8].
One of the most important differentiating traits between cultivated and wild grapevine is flower morphology [7,9]. Wild grapes have dioecious male and female plants, while cultivated varieties are mostly hermaphrodites. Dioecism in wild grapes is mentioned in ancient literature from the first century B.C. referring to grapevines with poor yields [10]. That wild grape, as the proposed ancestor of cultivated grape, underwent the dramatic change from dioecy to hermaphroditism is a primary hypothesis regarding grapevine domestication, and one of the most debated [11,12]. During domestication, it is hypothesized that the change from dioecious wild plants to hermaphrodite cultivated plants ensured greater yield and more stable, regular production [13].
Genetic relationships between wild and cultivated grapevines from Central Asia, the Mediterranean and Western Europe indicate considerable gene-flow and complicate differentiation of the two subspecies [14]. Admixed ancestry has been documented in most of research comparing wild and cultivated grapevines [14][15][16]. Gene flow evolutionary happened from wild ancestor to cultivated grapes but it is also possible from cultivated to wild grapevines, forming admixed individuals [17]. These two subspecies are reproductively compatible even though gene flow between them is slow (4-26%). It is linked to different flowering times in sylvestris and vinifera, and to geographic isolation of wild habitats [17][18][19]. Wild and cultivated grapevines form a genetic and taxonomic continuum, resulting in spontaneous hybrids [16]. Cultivated grapevine has greater overall genetic diversity and heterozygosity than wild [14,16,20] due to the fact that, during domestication, a large diversity was preferred (e.g., berry shape, aroma, yield) which ultimately favors heterozygosity. The grapevine was distributed very early following ancient trade routes that contributed to the development of a large number of varieties. There is evidence of first-and second-generation crosses between wild and cultivated individuals within the wild habitats [18]. Cultivated varieties clustered together with wild grapes from the same area in Georgia [14]. There was introgression of Italian wild grapes in the genome of some Italian cultivated varieties, in particular some local Tuscan varieties [20]. The relatively isolated Sicilian pool of wild populations was closely related to the local cultivated germplasm [9]. Clustering with local varieties was also observed in Portugal [16]. Around 70% of Iberian varieties are related to its sylvestris populations [10].
Grapevine classification according to Negrul [21] has been confirmed in many genetic studies [22][23][24], with grape varieties being classified into the following three main groups of proles: pontica, orientalis, and occidentalis. There are lower sub-classification levels for pontica ecotypes-georgica and balkanica.
Research on the presence and diversity of natural populations of wild grape in Croatia and Bosnia and Herzegovina is relatively recent [25]. Hungarian and German V. sylvestris populations were described earlier [3,26,27]. Neighboring countries such as Croatia, Bosnia and Herzegovina, Hungary, and Slovenia have had inter-related viticulture since ancient times [28,29], but there is very little data about genetic relationships among wild and cultivated grapevines within this region, or its connection with Western European germplasm. There are historical evidences that exchange of plant materials among distant European winegrowing regions was common. For example, the variety "Heunisch Weiss" (syn. "Gouais blanc") was most probably introduced to Germany and northwestern France from Eastern Europe [30]. "Heunisch Weiss" is a progenitor of many well-known European grapevine varieties such as Chardonnay and Riesling [31]. Other traditional cultivated varieties from Central Europe such as Traminer share close genetic similarity with V. sylvestris [32]. In Germany, it is most likely that selection of many natural seedlings from crosses between wild and cultivated grapes occurred in the Rhine floodplains, where these two subspecies coexisted together for a long time. There is also a presumption based on investigation of the Danubian wild grape population in Austria, the largest wild grape population in Europe, that it is the bastion of the former metapopulation [15]. Relatively high genetic diversity among true wild grapes in the Danubian populations suggested that the Balkan area is one of migrating populations, which contributed to the overall grapevine diversity of the Austrian populations [15]. The Danube river originates in Germany (Baden-Württemberg), passes through 10 different countries, and drains into the Black Sea. The Danube basin is one of the earliest human-settled areas in Europe.
The grapevine gene pools from Middle and Eastern Europe provide valuable insight into the broader context of this historically important area. The present dataset includes SSR genotyping data for 49 sylvestris samples from Croatia and 49 cultivated accessions that were previously described [25]. In Germany, more than 80 individuals are thriving in the most northern population of V. sylvestris on the island of Ketsch in the Rhine river [33]. The present study characterizes this genetic diversity from the Western Balkan Peninsula and Germany, including Hungary, to analyse its genetic structure and relationships among wild and cultivated grapes. These results will clarify grape taxonomy and aid grapevine breeders.

Plant Material
The present study is based on 243 non-redundant wild and cultivated grapevine genotypes from Croatia, Bosnia and Herzegovina, Slovenia, Hungary, and Germany (Supplemental Table S1). V. sylvestris samples were collected in their natural habitat: wet areas near rivers, lakes and other water reservoirs. Each sylvestris candidate was analyzed ampelographically and only those that met the basic dioecious phenotypic profile of wild grapevines were subjected to further genetic analysis. V. sylvestris indivuduals were sampled at their natural habitats on the eight different locations: Croatia (P01, P02, P04 and P06), Bosnia and Hercegovina (P03), Slovenia (P05), Hungary (P07), and Germany (P08) (Figure 1).
The KAPA Fast Multiplex PCR Kit (2x) (Kapa Biosystems, Wilmington, MA, USA) was used to set up reaction mixtures containing master mix, 100 pmol of each primer, and ~ 1 ng template DNA. Amplification was performed in ABI 9700 thermal cyclers (Applied Biosystems, Foster City, CA, USA) using the following program: three min initial denaturation at 95 °C, followed by 30 cycles of denaturation at 95 °C (15 s), annealing at 60 °C (30 s), and extension at 72 °C (30 s). A final extension was performed at 72 °C for seven min. DNA of two certified cultivars from Julius Kühn-Institut laboratory, "Muscat á petits grains" and "Cabernet franc," were amplified and used as references to standardize the allele calls.
The KAPA Fast Multiplex PCR Kit (2x) (Kapa Biosystems, Wilmington, MA, USA) was used to set up reaction mixtures containing master mix, 100 pmol of each primer, and~1 ng template DNA. Amplification was performed in ABI 9700 thermal cyclers (Applied Biosystems, Foster City, CA, USA) using the following program: three min initial denaturation at 95 • C, followed by 30 cycles of denaturation at 95 • C (15 s), annealing at 60 • C (30 s), and extension at 72 • C (30 s). A final extension was performed at 72 • C for seven min. DNA of two certified cultivars from Julius Kühn-Institut laboratory, "Muscat á petits grains" and "Cabernet franc", were amplified and used as references to standardize the allele calls.
Amplified products were resolved using capillary electrophoresis on an ABI 3130xl Genetic Analyzer (Applied Biosystems, Foster City, CA, USA) using GeneScan-LIZ 500 as an internal standard. Peaks were identified by size and height with GeneMapper 5.0 software (Applied Biosystems, Foster City, CA, USA).

Data Analysis
Different measures of genetic variability among 243 unique genotypes at 20 SSR loci were calculated. The number of alleles per locus (Na), number of effective alleles (Ne), private alleles, observed (Ho) and expected (He) heterozygosity, and Fixation index (F) were calculated for each locus over both wild populations and cultivated grapevines using GenAlEx 6.5 [42].
Genetic relationships among accessions were assessed by distance-based cluster analysis using the neighbor-joining method (NJ) as implemented in the MEGA 6.0 software [43] using the codominant genotypic distances between all pairwise combinations calculated by the GenAlEx 6.5 [42]. The bootstrap interior branch test implemented in MEGA 6.0 software [43] was used to test the reliability of each interior branch on tree. Principal coordinate analysis (PCA) was used to display genetic divergence among samples using codominant genotypic distances computed in GenAlEx 6.5 [42].
The genetic differentiation between populations was carried using pairwise F ST and pairwise Nei's genetic distance for each pairwise combination of populations in GenAlEx software [42].
Bayesian model-based cluster analysis executed in STRUCTURE [44] was applied to infer the genetic structure of the investigated accessions and to assign individuals to populations. The STRUCTURE configuration was set to ignore population information and use an admixture model with correlated allele frequencies. The degree of admixture α was inferred from the data. α is close to zero when most individuals are from one population or another, while α is >1 when most individuals are admixed. The allele frequency parameter (Lambda) was set to one as suggested in the STRUCTURE manual. Various numbers of putative populations (K) were tested, ranging from 1 to 10. Burning time and replication number were set to 100,000 and 100,000, respectively, in each independent run with 10 iterations. The choice of the most likely number of clusters (best K) was evaluated in accordance with the ad hoc statistic delta K as described [45] using Structure Harvester [46]. Structure bar plot was visualized by running the clumpp file (K = 2) obtained by Structure Harvester, in Structure Plot v 2.0 [47]. Structure ancestry Q values for each analysed individual are presented in Supplemental  Table S3, with the highlighted values of Q > 0.75 representing reliable ancestry assignment to its own cluster.

Results
The genetic indices of the 167 wild and 76 cultivated accessions at 20 nuclear SSR loci were calculated (Table 1) after the SSR profiles of all 243 analyzed samples were determined (Table S3). The total mean number of alleles (Na) was 11 and ranged from four (VVIQ52) to 19 (VMC4F3.1). The number of effective alleles (Ne) ranged from 1.957 (VVIQ52) to 6.512 (VVIV67), with an overall mean of 3.658. Observed and expected heterozygosity was least for VVIQ52; the largest Ho was for VVMD7 while the largest He was for VVIV67. VVMD7 is reported as the most informative marker in V. sylvestris [48]. The expected heterozygosity (He) values varied between 0.489 (VVIQ52) and 0.846 (VVIV67), with an average of 0.698. Overall mean observed heterozigosity was lower than the expected heterozygosity. The observed heterozygosity (Ho) values ranged from 0.364 (VVIQ52) to 0.755 (VVMD7), with an overall mean of 0.578. Fixation index (F), a measure of reduction in heterozygotes and thus an indicator of greater inbreeding, was least for VVMD7 (0.047) and greatest for VVMD21 (0.340). The mean F was 0.174. Genetic diversity indices at the population level show that the number of alleles per locus of cultivated grapevines was greater (8.600) than in wild grapevines, where Na values ranged from 3.100 (Hungarian wild set = P07) to 5.300 (Bosnian wild set = P03) ( Table 2). The overall Ne value in the dataset was 2.739. Cultivated varieties had the greatest (4.471) Ne values, while P07 had the lowest value (2.000) and the Croatian P04, the highest (2.796) among the wild populations. The Ho values in vinifera (0.759) were greater than that of any wild population, which ranged from 0.437 in P08 to 0.593 in P06. The He values were greater for vinifera grapevines than for sylvestris. Fixation index values were positive for all analysed groups except for cultivated varieties, indicating heterozygosity deficiency or greater inbreeding within wild populations. The group of cultivars had a negative F, indicating more heterozygosity. The F values within sylvestris were least for P07 (0.013) and greatest for P05 (0.146) Table 2. Genetic diversity estimates for each analyzed population of wild and cultivated grapevines. Population P01 was excluded due to an insufficient number of individuals (only two) for population genetic indices. Genetic distance among the eight sylvestris populations and the group of cultivated varieties was estimated using pairwise Fst values and Nei's genetic distance. (Table 3). The Nei's genetic distance values ranged from 0.039 (Croatia-Bosnia sylvestris, P01-P03) to 0.625 (cultivated group-Germany sylvestris P08). The F ST values ranged from 0.067 for the Croatian sylvestris (P04) and Slovenian sylvestris (P05) up to 0.163 for the sylvestris from Croatia (P01 and P02) but was greatest for sylvestris P01-cultivated samples (0.187). Fifty-seven private alleles (PA) were found at 19 out of 20 SSR markers in the studied set (Supplemental Table S4). In sylvestris, PA were observed at 15 out of 20 SSR. The number of PA was lower in sylvestris (22) than in the cultivated vinifera (35), but not drastically so. In cultivated grapevines, PA were found at 14 out of 20 SSR loci. In this study, 22 sylvestris PA were observed in seven out of eight populations. The greatest number of private alleles was identified in P06 (6) followed by P05 (5), P03 (4), and P08 (3), all populations from different countries.
The neighbor-joining (NJ) cluster analysis showed clear differentiation between the vinifera and sylvestris subspecies, which formed two main clusters, one of which contained cultivated grapes (C1), and other (C2), the wild grapes from Germany, Hungary, Slovenia, Croatia, and Bosnia and Herzegovina ( Figure 2). One set of Western European wine cultivars ("Riesling blau", "Burgunder gross", "Moehrchen", "Pinot meunier", "Chatus", "Heunisch blau", "Adelfraenkisch", "Hartblau", "Gewuerztraminer", and "Ondenc") showed affinity for the wild grapes, forming the basal group to the rest of the cultivated grape. There are also a few sylvestris individuals from various wild populations (P07_7, P05_2, P03_1, P03_5, and P03_6) that are closely allied with the cultivars. Wild accessions tend to cluster together with individuals from the same population due to high within-population genetic similary. The geographic clustering can be seen in the vinifera set originating from at least 10 different countries tending to differentiate between Western and Eeastern European cultivars. The structure and the correlations among individuals were analysed using PCooA based on the genetic distance matrix of 20 SSRs. PCooA projections were plotted in a two-dimensional scatter plot ( Figure 3). The PCooA maximizes the linear correlation between distance in the distance matrix and distance in the 2-D space, giving insight into the relative relationships among accessions. Projection of the first two principal axes accounted for 23.27% of the total molecular variation. The first dimension (PCooA1) explained 17.02% while the second dimension (PCooA2) explained 6.25% of the total variation in the set. Genetically similar individuals were highly correlated and clustered together, forming two groups (cultivars with cultivars-C1 and sylvestris with sylvestris-C2). Despite the strict pre-analysis selection of presumed pure sylvestris, PCooA revealed slight overlap between C1 (vinifera) and C2 (sylvestris), indicating gene flow from vinifera to sylvestris and vice versa, even though the vinifera set was half the size of the wild set. The structure and the correlations among individuals were analysed using PCooA based on the genetic distance matrix of 20 SSRs. PCooA projections were plotted in a two-dimensional scatter plot (Figure 3). The PCooA maximizes the linear correlation between distance in the distance matrix and distance in the 2-D space, giving insight into the relative relationships among accessions. Projection of the first two principal axes accounted for 23.27% of the total molecular variation. The first dimension (PCooA1) explained 17.02% while the second dimension (PCooA2) explained 6.25% of the total variation in the set. Genetically similar individuals were highly correlated and clustered together, forming two groups (cultivars with cultivars-C1 and sylvestris with sylvestris-C2). Despite the strict pre-analysis selection of presumed pure sylvestris, PCooA revealed slight overlap between C1 (vinifera) and C2 (sylvestris), indicating gene flow from vinifera to sylvestris and vice versa, even though the vinifera set was half the size of the wild set. Non-hierarchical horizontal clustering with the Structure software assigned the 243 specimens (167 V. sylvestris and 76 V. vinifera) into two clusters: sylvestris and vinifera (Figure 4) based on the optimal number of clusters (K) calculated as described [46] (Supplemental Figure S1). Both groups were clearly separate, but also showed some admixture, corresponding to the overlap observed in the NJ and PCooA analysis. The accessions with Q < 75 were considered accessions with admixed origin. All populations except P05 (Slovenia) showed admixed sylvestris/vinifera ancestry. Among wild populations, the most accessions of admixed ancestry were observed in Croatian populations P06 (Psunj) and P02 (Imotski). Populations P07 (Hungary) and P08 (Germany) had only one such accession. The greatest percentage of admixed genomes among cultivated V. vinifera was observed in Western wine cultivars: "Pinot meunier," "Burgunder gross," "Moehrchen," "Suessschwarz," "Adelfraenkisch," "Gewuerztraminer," "Riesling blau," "Chatus," and "Ondenc." Other cultivars were all of V. vinifera ancestry with Q > 75 (Supplemental Table S5). Structure showed similar clustering to NJ and PCooA. However, NJ and Structure were more subtle, providing more detailed insight into the genetic structure among individuals than the distance-based PCooA. Detection of intermediate position between the groups (overlaps) on the graph is neverthless facilitated by PCooA, but the multiple dimensions to show the relationship between genotypes provided by clustering methods.  Non-hierarchical horizontal clustering with the Structure software assigned the 243 specimens (167 V. sylvestris and 76 V. vinifera) into two clusters: sylvestris and vinifera (Figure 4) based on the optimal number of clusters (K) calculated as described [46] (Supplemental Figure S1). Both groups were clearly separate, but also showed some admixture, corresponding to the overlap observed in the NJ and PCooA analysis. The accessions with Q < 75 were considered accessions with admixed origin. All populations except P05 (Slovenia) showed admixed sylvestris/vinifera ancestry. Among wild populations, the most accessions of admixed ancestry were observed in Croatian populations P06 (Psunj) and P02 (Imotski). Populations P07 (Hungary) and P08 (Germany) had only one such accession. The greatest percentage of admixed genomes among cultivated V. vinifera was observed in Western wine cultivars: "Pinot meunier", "Burgunder gross", "Moehrchen", "Suessschwarz", "Adelfraenkisch", "Gewuerztraminer", "Riesling blau", "Chatus", and "Ondenc." Other cultivars were all of V. vinifera ancestry with Q > 75 (Supplemental Table S5). Structure showed similar clustering to NJ and PCooA. However, NJ and Structure were more subtle, providing more detailed insight into the genetic structure among individuals than the distance-based PCooA. Detection of intermediate position between the groups (overlaps) on the graph is neverthless facilitated by PCooA, but the multiple dimensions to show the relationship between genotypes provided by clustering methods.
1 Figure 4. Graphic presentation of the population structure of 243 grapevine accessions. Each accession is represented by a single vertical bar divided into K color segments representing its proportions in the two inferred genetic clusters using STRUCTURE software. Wild accessions grouped into a population represented by green, while cultivated accessions grouped into a population represented by blue.

Disscussion
The western Balkans and Central Europe were primary transit routes of grapevine varieties on their way from the east to the west [23,49]. The main objective of this study was to investigate genetic diversity, relationships and structure between wild and putative autochthonous varieties in this area to find whether their common territorial and historical ground is reflected in viticulture, and to describe the wild gene pool in relation to the cultivated pool of this area.
Genetic diversity in the studied set of individuals, expressed through the expected and observed heterozygosities of SSR markers, was greater in vinifera than in sylvestris, as seen in previous studies [9,16,18,20]. Among sylvestris populations, the greatest heterozygosity was observed in P06 (Croatia) and the lowest for P08 (Germany), indicating heterozygosity decreased from east to west. The east-west pattern of decreasing diversity was documented in other studies [4,24]. It is presumed that sylvestris populations in the Rhine floodplains followed a south-north migration route after glaciation retreated, being thus more related to the Southwestern European gene pool. Decreasing heterozygosity was confirmed by the fixation index, in general highlighting inbreeding in all sylvestris populations, while the cultivated set had a negative fixation index, showing greater genetic diversity than sylvestris. However, this situation is not expected in sylvestris, as the presumed ancestor of cultivated grapes. The F ST statistics showed low differentiation between vinifera and sylvestris, in agreement with previous results [14,49]. It was greater between sylvestris-sylvestris (Croatia-Slovenia), but even greater between sylvestris (Croatia, P01) and cultivated varieties. Genetic diversity by all measures was reduced in sylvestris populations, indicating their isolation, where sibmating leads to inbreeding and loss of alleles and heterozygosity. Private alleles can be useful indicators of gene flow [50] or used as discriminant markers to reliably distinguish sylvestris from vinifera [23]. In this study, while overall private allelic richness was lower in sylvestris than in the cultivated set, there were many private alleles in the sylvestris populations from all countries, highlighting the local richness of the wild populations. There was continuous reduction in genetic variability of Vitis sylvestris populations, leading to loss of alleles and heterozygosity. In general, perennial fruit crops such as grape are the descendants of spontaneous or controlled crosses between geographically and genetically distant individuals, resulting in new variants carring novel combinations of alleles not find in the wild [51]. Therefore, cultivars contain more genetic variability than the remaining wild grape populations. Unique alleles observed in only one population or individual can be useful in determining to which population a particular individual belongs, as indicators of gene flow [50].
By using the combined vinifera -sylvestris set, the NJ, PCooA, and STRUCTURE analyses confirmed clear separation of the two groups. Both groups showed high assignment to their own cluster, confirming the representativeness of the studied set. The accessions falling in the transition zone in NJ, the overlapping zone in PCooA or the admixed part in Structure suggest a common gene pool for the two groups. Despite the strict pre-selection of true-to-type sylvestris individuals and a cultivated set originating from at least 10 different countries that was only half as large, classification (hierarchical, nonhierarchical) and ordination (PCooA) methods both showed bidirectional gene flow between the vinifera and sylvestris gene pools.
The NJ cluster analysis was carried out without considering the geographic origin of the samples. However, the cultivated set displays a clear structured arrangement of accessions by ancestry, diversity and putative geographic origin. "Elbling Blau" and eight wine grapes in the first three independent small subclusters are first-degree related to 'Savagnin' (specifically, its red mutant 'Traminer') and 'Pinot'. The genetic proximity of "Savagnin"/"Traminer" and "Pinot" to sylvestris shown here was reported previously by various authors [1,14,49]. All but two varieties in the subsequent cluster comprising 12 varieties are first-degree related to "Heunisch", with one variety being first-and four varieties being second-degree related to "Savagnin"/"Traminer." Genetic composition of "Riesling Weiss", "Elbling", and "Raeuschling" and morphology place them close to "Savagnin"/"Traminer." The presence of alleles of the at least 2000-year-old "Savagnin" [52] in the subgroup possibly explains the position near to the sylvestris cluster. Most of these wine grapes, except "Riesling Weiss", "Pinot", and "Gewürztraminer", are minor varieties or were rediscovered in recent times. The subsequent clusters encompass mainly Hungarian varieties, followed by Croatian and Bosnian with some intermixed Hungarian varieties. Clustering far from sylvestris and high vinifera ancestry values point to a different geographic origin, most likely the Near East [50]. In the wild cluster, the structure is fragmented but follows the geographic pattern of the populations, confirming high similarity among wild populations and their common gene pool. The Q values indicate admixed individuals in all wild populations. The Croatian population from Psunj (P06) had the most admixed individuals (7), suggesting more recent introgression from cultivated grapes, while in other wild populations, such individuals were rarer. Wild accessions from Germany and Hungary were far less likely to be assigned to vinifera cluster, while cultivars from Germany were grouped close to the sylvestris cluster. "Riesling blau", the cultivar in the transitional zone in NJ, shared more alleles with sylvestris than with the vinifera group in this set of samples (Q: sylvestris = 0.69, vinifera= 0.31). The same pattern of allele sharing from sylvestris was seen in some other Western European cultivars in other genetic comparison studies [14,49,50]. When analysed using SSRs, the cultivar "Manseng" clustered with sylvestris in a Structure plot [23], but SNP analysis clustered it among vinifera. One should take into consideration the limited power of SSR markers to resolve subtle phylogenetical relations among closely related individauls [23,53].
Because of its reduced number of dimensions, PCooA provides less-detailed insight into the genetic structure of the studied set and does not allow conclusions as to what extent some population/individual influenced the other. However, this analysis facilitates detection of an intermediate position between the groups (overlaps). Clustering methods give more detailed information on the genomic nature of the studied set. Both groups showed very high average probabilities of assignment to their own cluster, consistent with being genetically different.
Outlier groupings or diversity was not observed, giving insight into the similarity of the studied wild populations from this area. Greater diversity was observed in the eastern than in the western wild populations, and southern populations were more diverse than the northern ones, so the decrease in diversity can be traced from east to west, and from south to north. Ketch island in Germany was the northernmost population, while Dalmatian populations were most southern. In the Balkans, one study assumed there was a past metapopulation covering the broader area of Danube river [15]. Several populations from different countries belong to this area. All studied Vitis sylvestris populations represent small, isolated populations surviving in mountain or floodplain forests in close proximity to rivers. The remaing populations of Vitis sylvestris are considered an "endangered species" [15]. Several sites (P02, P04, P07 and P08) belong to protected areas of flora and fauna biodiversity, where no direct human degradation is allowed, but all sites are endagenred due to taxonomic pollution through gene flow from other Vitis taxa that escaped from vineyards [15]. Gizdavac (P01) is a highly endangered wild population with only two individuals (female and male) recorded.
This study demonstrated that populations of wild grapes are highly fragmented into small, isolated populations that represent valuable genetic resources. Additional conservation efforts are needed to protect the remaining genetic variation of V. v. subsp. sylvestris resources.

Conclusions
Analysis of Western Balkan and Central European V. vinifera and V. sylvestris populations showed clear distinction between cultivated and wild grapevines. Despite strict selection of pure sylvestris and a considerably smaller set of vinifera cultivars included in analysis, all statistical and phylogenetic methods showed high assignment of vinifera to vinifera and sylvestris to sylvestris clusters, but always with visible overlaps between the groups. Through hierarchical phylogenetical clustering, vinifera cultivars were shown to be the source of genetic variation in the studied set. Allelic richness of the wild gene pool from various countries indicated local adaptation of sylvestris populations and assignment to its population of origin, confirming a valuable source of genetic variation in the studied populations. Close affinity was seen between the studied sylvestris populations and Western European cultivars, suggesting a common genetic origin. In contrast, cultivated grapes from the Balkan area were distant from the sylvestris populations in this sample set, suggesting a different geographic and genetic pool of domestication. However, the general questions relating the time and direction of introgression between these two subspecies are still open and could be resolved with a broader sample set and whole-genome data.
Supplementary Materials: The following are available online at http://www.mdpi.com/2073-4425/11/9/962/s1. Table S1: List of samples used in this study, Table S2: characteristics of 20 SSR markers used in this study, Table S3: SSR profiles of all 243 analyzed samples, Table S4: List of private alleles (and allele frequencies) in a 243-sample set of V. sylvestris and V. vinifera, Table S5: Q values of 243 samples obtained from Structure software analysis, Figure S1: Bayesian assignment analysis as: (a) a plot of mean likelihood L(K) and variance per K value exploited by STRUCTURE software on a data set of 243 individuals based on 20 SSR loci; (b) The optimal number of K that best fit the data set is two.