2.1 Seed material

The seeds of 18 black berry and 22 white berry Sardinian cultivars, among the 61 studied by De Mattia et al. [12], native of the whole Sardinian region, but grown in the Agris germplasm collections (Agenzia per la Ricerca in Agricoltura della Regione Sardegna) of Ussana (Sardinia, Italy), were collected during the harvest years 2008–2009 and 2009–2010, for a total of 40 seed accessions of grapevine (Table 1 and Fig. 1). Grapes have been collected at the time of maximum concentration of sugar in the pulp (ripeness) corresponding to the complete morphologic and chromatic seed development. For each cultivar, 10 bunches were collected from 10 different individuals and from the central part of each bunch, a variable number of berries (from 10 to 30), depending on bunch and barriers morphologic characteristics, were collected. Undeveloped seeds (stenospermocarpic) have been not considered [17].

Table 1

Name code, grape variety name, colour (B: black; W: white), distribution and number of seeds of V. vinifera L. studied cultivars.

Name code	Grape variety	Grape colour	Distribution	Amount of analysed seeds
ArB	Aregu Biancu	W	Barbagia di Seùlo	600
ArG	Aregu Giallo	W	Barbagia di Seùlo	598
ArM	Argu Mannu	W	Sardinia	1192
Cn	Cannonau	B	Barbagia di Seùlo, Ollolai e Belvì	1776
CnO	Cannonau Bianco di Oliena	W	Barbagia di Ollolai	603
CnT	Cannonau Bianco di Trieri	W	Ogliastra	1200
Cl	Culupuntu	W	Ogliastra	601
FlG	Falso Gregu	B	Campidano di Cagliari	595
GB	Gregu Bianco	W	Campidano di Cagliari e Oristano	1049
GNC	Gregu Nieddu del Campidano	B	Campidano di Cagliari e Oristano	1199
GNS	Gregu Nieddu di Serramanna	B	Campidano di Cagliari	598
Mz	Manzesu	B	Campidano di Cagliari	2098
ME	Monica di Escalaplano	B	Sarrabus e Gerrei	600
MSl	Monica di Seulo	B	Barbagia di Seùlo	596
MSr	Monica di Sorgono	B	Mandrolisai	600
MLd	Moscato di Lodine	W	Barbagia di Ollolai	596
MPt	Moscato di Pattada	W	Monteacuto	600
MT1	Moscato di Tempio 1	W	Gallura	597
MT2	Moscato di Tempio 2	W	Gallura	600
MSP	Moscato Su Pinu	W	Barbagia di Ollolai	600
NrT	Nera Tomentosa	B	Planargia	599
NrB	Nera di Bosa	B	Planargia	600
NPd	Nieddu Mannu di Padria	B	Meilogu	596
NPt	Nieddu Mannu di Pattada	B	Monteacuto	595
NPS	Nieddu Pedra Serra	B	Campidano di Sanluri	599
NPl	Nieddu Polchino	B	Monteacuto	596
Nr	Nuragus	W	Campidano di Cagliari	888
NrA	Nuragus Arrubiu	W	Campidano di Cagliari	598
Nrd	Nuragus Muscadeddu	W	Sarrabus e Gerrei	600
Nrt	Nuragus Moscatello	W	Sarcidano	600
NRR	Nuragus Rosso Rompizzolla	W	Campidano di Cagliari	599
PsC	Pascale di Caglairi	B	Campidano di Cagliari	1766
PsO	Pascale di Oliena	B	Barbagia di Ollolai	599
PrN	Primidivu Nieddu	B	Meilogu	596
Sn	Sinnidanu	W	Baronie	1193
Vr	Vernaccia	W	Campidano	1193
VrE	Vernaccia di Escalaplano	W	Sarrabus e Gerrei	608
VrR	Vernaccia di S. Rosalia	W	Ogliastra	597
VrS	Vernaccia di Solarussa	W	Campidano di Oristano	1049
Vrt	Vertudi	B	Sulcis	1200

2.2 Seed analysis

2.3 Statistical analysis

Using the same approach that has been many times applied to solve cases of homonymy and synonymy on the basis of molecular markers, morpho-colorimetric features were used to identify hypothetic grapevine cultivar synonymy groups. A total of 113 morpho-colorimetric features was used to characterize the germplasm of the 40 studied Sardinian grapevine cultivars (Table 2).

In order to compare the results of genetic analysis achieved by De Mattia et al. [12] on Sardinian grapevine cultivars with the relative seed phenotypic expression, a preliminary morpho-colorimetric comparison among 40 of the 61 cultivars tested by De Mattia et al. [12] was executed on the basis of the 13 synonymy groups proposed by De Mattia et al. [12]. Table 3 shows the results of this preliminary comparison. Although the overall discrimination performance only reached 59.7%, because of the non-inclusion in any group of the 15 cultivars proposed by De Mattia et al. [12], the percentages of correct identification of each group show a certain correspondence to the phenotypic seed characters (Table 3), simultaneously suggesting some potential modifications to the proposed groupings.

From two separate comparisons among black and white berry cultivars (Tables 4 and 5), only considering the varietal name as grouping variable and excluding the synonymy group notations of De Mattia et al. [12], evaluating the percentages of correct identification and above all the mistakes made by the classification system, it was possible to identify new hypothetical synonymy groups. Tables 4 and 5 show the results of the comparisons among the analysed black berry and white berry cultivars, respectively. In the first case, an overall percentage of correct discrimination of 58.8% was reached, while white berry cultivars were correctly identified in the 60.1% of the cases. Figs. 2 and 3 show the graphical representations of the achieved groupings from the morpho-colorimetric comparisons among the black berry and white berry cultivars, respectively. Although the 3D plots were drawn using only three of the about 20 discriminant functions implemented to distinguish the varieties in both the cases, these graphical representations well-highlight the phenotypical differences among the membership groups, organizing little clearly identifiable groups. The three discriminant functions used to draw the plots were able to explain the 66.3% and the 70.7% of the whole statistical variability for the black berry and white berry cultivars comparisons, respectively (data not shown).

Following the suggestions of the implemented classifiers, on the basis of the achieved results by the two discriminant statistical elaborations, two further discrimination analyses were carried out, moving some of cultivars from a group to another and hypothesizing new synonymy groups. It was possible to identify six black berry cultivar synonymy groups (Table 6) and eight white berry cultivar synonymy groups (Table 7). The black berry cultivar Manzesu (Mz) was moved from the group 2 (G2) proposed by De Mattia et al. [12], where it was considered synonym of Monica di Escalaplano (ME) and Monica di Seulo (MSl), to the G3 together with Monica di Sorgono (MSr) and Nieddu Mannu di Padria (NPd). The displacement of this black berry cultivar, brought to the increment of the percentages of correct identification of the two groups from 53.4% (Table 3) to 68.6% (Table 6) and 43.1% (Table 3) to 77.4% (Table 6) for G2 and G3 respectively (Fig. 2). According to De Mattia et al. [12], considering the wide diffusion of the Monica cultivars in Sardinia, the presence of the variety known as Monica di Sorgono (MSr) into the G3 rather than into the G2 together with the other Monica cultivars, suggests the possibility that a wrong name was historically attributed to this cultivar. Moreover, this modification allows one to suppose that probably the cultivar named Manzesu (Mz) should be the same variety of those currently known as Monica di Sorgono (MSr) and Nieddu Mannu di Padria (NPd) rather than of Monica di Escalaplano (ME) and Monica di Seulo (MSl) as reported by De Mattia et al. [12]. This supposition is also graphically supported by the spatial distance of the barycentre discriminant scores of these cultivars (Fig. 2).

Moreover, the white berry cultivars Nuragus (Nr) and Nuragus Arrubiu (NrA), considered as independent by De Mattia et al. [12], were moved into the G4 together with Nuragus Moscatello (Nrt), Nuragus Rosso Rompizzolla (NrR) and Nuragus Moscadeddu (Nrd). Similarly to the previous case, this modification gave a rise in the performance from 69.5% (Table 3) to 80.4% (Table 7) for this synonymy group (Fig. 3).

Also, the independent white berry cultivar Vernaccia di Escalaplano (VrE) was moved into the G5 together with Vernaccia di S. Rosalia (VrR), Vernaccia (Vr) and Aregu Biancu (ArB), producing the same effect on this synonymy group which percentage of correct identification increased from 76.9% (Table 3) to 90.8% (Table 7), allowing one to advance the hypothesis that these four grapevine varieties are the same cultivar (Fig. 3).

The G7, constituted by Cannonau bianco di Triei (CnT) and Cannonau bianco di Oliena (CnO) was enriched with Vernaccia di Solarussa (VrS) and Sinnidanu (Sn), both considered as independent by De Mattia et al. [12]. This arrangement shown an improvement of the discrimination performance, bringing to the increment of the percentages of correct identification of this group from 62.2% (Table 3) to 85.2% (Table 7).

Another interesting case of potential synonymy is represented by Moscato di Tempio 1 (MT1), Moscato di Lodine (MLd) and Moscato Su Pinu (MSP) (Fig. 3). Although these three cultivars were considered different by De Mattia et al. [12] even if genetically close one to the other, here they are included in a new synonymy group (here labelled as G10) on the basis of germplasm phenotypic characters. This group shows a correct identification percentage of equal to 87.5% (Table 7).

Similar considerations for the white berry cultivars Culupuntu (Cl), Moscato di Tempio 2 (MT2) and Moscato di Pattada (MPt) (here labelled as G9). The 79.2% of correct identification obtained for this group (Table 7), as well as the graphical representation of the spatial distance of the discriminant scores (Fig. 3) should confirm the advanced supposition.

The same observations were done for the independent black berry cultivars known as Vertudi (Vrt), Gregu Nieddu del Campidano (GNC) and Gregu Nieddu di Serramanna (GNS) (Tables 3 and 4). The new synonymy G14 was carried out, reaching 75.0% of correct identification (Table 6) and confirming the possibility that these three cultivars are the same (Fig. 2).

Finally, according to the findings of De Mattia et al. [12], the white berry cultivars Gregu Bianco (GB), Aregu Giallo (ArG) and Argu Mannu (ArM) (here labelled as G11, G12 and G13, respectively) seems to be phenotypically and genotipically different when compared with the other studied cultivars, showing percentages of correct discrimination included between 64.3% and 85.8% (Table 7 and Fig. 3).

Considering the new synonymy groupings here proposed, the overall percentages of correct identification risen up to 81.0% for the six black berry cultivar groups (Table 6) and 82.8% for the eight white berry cultivar groups (Table 7).

Table 8 shows the comparison between the synonymy groups proposed by De Mattia et al. [12] on the basis molecular data and the new hypothetical synonymy groups achieved on the basis of morpho-colorimetrical data. It is possible to notice that five of the 14 groups remained unchanged (G1, G6, G8, G12 and G13), five of the studied varieties were unified to groups proposed by De Mattia et al. [12] (Nuragus and Nuragus Arrubiu in G4, Vernaccia di Escalaplano in G5 and Sinnidanu and Vernaccia di Solarussa in G7), tree groups were modified (G2, G3 and G11) and three new groups were formed (G9, G10 and G14).

As discussed above, Sardinian grapevines are characterized by a huge number of cultivars [6,7,11]. Some of these surely derive from different breeding events involving both local and out-coming material, as well as both domesticated and wild grapes, but despite the large and complex grapevine survey, many Sardinian varieties simply are the product of linguistic distortion, due to the wide historic-cultural heterogeneity of the island. This phenomenon generated a great assortment of grape names, that, together with the huge real number of cultivars, is the cause of the incredible current grapevine Sardinian panorama [12].

Many times molecular methods proved to be able to screen biodiversity genotyping and comparing different plant cultivars, allowing direct inferences about genetic diversity and interrelationships among organisms at the DNA level without the confounding effects of the environment and/or faulty pedigree records. Nevertheless, considering that the various molecular marker techniques are based on different principles, the importance of finding the right compromise between reliability and ease of analysis, statistical power and confidence of revealing polymorphisms, is extreme, and the solutions found not always satisfactory [32]. Moreover, frequently costs and execution times of these techniques are the undoubtedly disadvantageous [33]. Furthermore, it is important to consider a few inherent problems associated with the use of SSR markers. Sometimes, homoplasy is a phenomenon observed when a character present in two species is not derived from a common ancestry but rather, the similarity is a result of convergence, parallelism or reversion, so different copies of a locus are identical in state despite not being identical by descent [34]. Other times, in SSR analyses of a large number of samples from diverse germplasm, a few samples fail to produce PCR products, causing the impossibility to determine whether the absence of PCR products represents true null alleles of the SSR locus or is due to a failure of the PCR reaction [34]. Finally, depending from genome samples, the molecular analysis of restricted sequences and few alleles is not enough to discriminate among different cultivars, and the analysed alleles are not always related to varietal characters. For this reason, the potentialities of computer vision as complementary to the molecular methods, were here presented, highlighting how a wide gamma of phenotypic characters can be a valid choice to evaluate differences among cultivars also in a so complex scenario.

In the case of the synonymy study of grapevine cultivars, the 113 measured morpho-colorimetric features of the germplasm resulted enough to achieve a clear discrimination among the synonymy groups, as confirmed by the previous SSR analysis conducted by De Mattia et al. [12] on the same material. Finally, it could be very interesting to assess the contribute of a supporting ampelographic characterization, in particular using image analysis techniques on grapevine leaves, in the varietal identification, as well as in the synonyms and false attribution evaluations.

The authors declare that they have no conflicts of interest concerning this article.