Author Correction: Unraveling the genetic origin of ‘Glera’, ‘Ribolla Gialla’ and other autochthonous grapevine varieties from Friuli Venezia Giulia (northeastern Italy)

‘Glera’ and ‘Ribolla Gialla’ are the most economically relevant local grapevine cultivars of Friuli Venezia Giulia region (north-eastern Italy). ‘Glera’ is used to produce the world-renowned Prosecco wine. ‘Ribolla Gialla’ cultivation is constantly increasing due to the strong demand for sparkling wine and is the most important variety in Brda (Slovenia). Knowledge of local varieties history in terms of migration and pedigree relationships has scientific and marketing appeal. Following prospections, genotyping and ampelographic characterization of minor germplasm in Friuli Venezia Giulia, a further research was developed to understand the parentage relationships among the grapevine varieties grown in this region. An integrated strategy was followed combining the analysis of nuclear and chloroplast microsatellites with the Vitis 18k SNP chip. Two main recurrent parents were found, which can be regarded as “founders”: ‘Vulpea’, an Austrian variety parent-offspring related with at least ten Friuli Venezia Giulia cultivars, among them ‘Glera’, and ‘Refosco Nostrano’, first degree related with other six Friuli Venezia Giulia varieties. ‘Ribolla Gialla’ was shown to be another member of the impressively long list of offspring derived from the prolific ‘Heunisch Weiss’. Combining molecular markers and historical references was a high-performance strategy for retracing and adjusting the history of cultivars.

ASSIsT (Automatic SNP ScorIng Tool) software 23 v. 1.02 was applied for additional SNPs pruning and improved clustering classification of SNPs previously selected using GenomeStudio tools. The aim was to better evaluate pedigree results on putative PO related cultivars conflicting with previous results reported in the literature. ASSIsT default parameters were applied; no SNP map position was given, because ASSIsT does not work with random or unmapped SNPs. The pedigree file was given with mother and father information missing, also for the pedigrees used as references. After pruning, the selected SNPs were analyzed again on a larger set of 192 genotypes to obtain a more consistent SNP classification into the four classes elaborated by ASSIsT: Robust, OneHomozygRare_HWE, OneHomozygRare_notHWE, DistortedAndUnexpSegreg. The more restricted set of higher-quality SNPs was used to recalculate Mendelian inconsistencies on putative PO-related cultivars.
Full-sib relationships were inferred using Colony software version 2.0.6.5 (July 30,2018), freely available at https://www.zsl.org/science/research-projects/software. Empirical data input consisted of both nSSR and SNP markers. The Parent-Parent-Offspring trios inferred with PLINK and GenomeStudio were used to define the main structure of the family into Colony. The following main settings were applied: markers error rate 0.00001, no sibship prior indicator, one medium run, FL (full likelihood) analysis method and medium precision in calculating FL. Only full-sib dyads with probability equal to 1 were retained. Colony reconstructs the possible genotypes of the parents absent in the dataset and complementing the full-sibs found; these genotypes were used to evaluate additional PO relationships suggested by the software.
MEGA X software version 10.0.5 24 was used to produce an unrooted dendrogram of genetic similarity using the first-round 11,929 selected SNP markers. Pairwise genetic distances were computed using the Kimura 2-parameter method. Missing data were removed for each sequence pair (pairwise deletion option). The dendrogram was constructed using the Unweighted Pair-Group Arithmetic Average Method (UPGMA). A bootstrap test of 500 replicates was used to define the percentage of replicate trees in which the associated genotypes clustered together; these values were shown next to the branches. Only branches with bootstrap values higher than 75 were taken into consideration.

Results
The pedigree relationships of FVG grapevine varieties and others from neighboring countries were investigated by a well-established low-throughput SSR approach and a high-throughput genotyping system based on an SNP chip array, the Vitis 18k SNP.
SSR genotypes and ampelographic features of most varieties of interest for FVG are available on-line in the Italian Vitis database (http://www.vitisdb.it/). VIVC prime names and variety numbers for FVG varieties are reported in Table S1. If the prime name differs considerably, it is added in brackets in Table S2, which reports chlorotypes and nSSR profiles of all studied varieties.
Filtering reduced the initial 18,071 SNPs available in the chip of ILLUMINA to 11,929. All varieties were univocally identifiable with the 14 SNP set selected by 25 ; these profiles are provided in Table S3. The SNP complete profiles are in Table S4.
Taking the 'Heunisch' samples and related synonyms ('Liseiret' and 'Rebula Stara'), 18 out of 11,929 SNPs failed, all others showed the same allelic combination except one, 'Heunisch Weiss Seedless' (data not shown). Concerning the two 'Ribolla Gialla' samples, one from the Italian Collio and one from the Slovenian Brda, only 5 SNPs were missing, and the others showed the same allelic profile.
'Marzemina Bianca' and 'Codelunghe' were shown to be PO related to 'Garganega' , like 'Brambana' to 'Bianchetta Trevigiana'; moreover, 'Ruacit' and 'Blanchias (Blancjàs)' could also be FS related (Fig. 1). (2020) 10:7206 | https://doi.org/10.1038/s41598-020-64061-w www.nature.com/scientificreports www.nature.com/scientificreports/ Some pedigrees derived from previous SSR studies were not supported by the SNP chip analysis, because PLINK parameters and Mendelian inconsistencies were incompatible with PO relationships compared to the references ( Table 2). In more detail, Z1 value was expected to be not less than 0.8279, given the references used, instead, the values found were too low and only between 0.5035 and 0.6188. Regarding Mendelian inconsistencies (MI), PO relationships given as references are 0.0021 for Pinot and Traminer, and 0.0031 for Traminer and Sauvignon Blanc; MI showed to be a little bit higher for Riesling Weiss and Heunisch (0.0037), a pair of varieties already assumed to be PO related by the literature data (VIVC). MI lower than or very close to the highest value for the reference pairs (0.0037) were computed for the pairs of varieties considered to be PO related; instead, MI 2 to 9 times higher than the highest value of reference were computed for pairs excluded as being PO related. Specifically, our data reject the PO relationships for 'Marzemino' and 'Teroldego' and for 'Marzemino' and 'Refosco dal Peduncolo Rosso' supported by other studies 8,27 . Furthermore, SNP data do not support the PO relationship for 'Marzemino' and 'Marzemina Bianca' 28 nor for 'Glera Lunga' and 'Glera' 8 . Instead, Colony output supported the hypothesis that 'Marzemino' and 'Refosco dal Peduncolo Rosso' could be FS related (Fig. 2).
ASSIsT software was applied to the set of 11,929 SNPs previously selected using GenomeStudio tools: 9,177 SNPs (76.93%) were retained and more consistently classified using a larger set of 192 genotypes. The 9,177 SNPs were used for Mendelian inconsistencies (MI) computation. The obtained results are reported in Table 2 www.nature.com/scientificreports www.nature.com/scientificreports/ from 9 to 33, among them Riesling Weiss with 31 total MI. Varieties with conflicting data showed higher total MI, from 88 to 386, than varieties with supporting data, which is at least 2.7 times higher than supported PO relationships.
MI were also split in three classes: Robust, DistortedAndUnexpSegreg, and OneHomozygousRare (encompassing both OneHomozygousRare_HWE and OneHomozygousRare_not HWE) ( Table 2). Given that MI based on DistortedAndUnexpSegreg can be ascribed to AB × AO marker in germplasm population, Robust and OneHomozygousRare based MI were the strongest reference values for PO relationships evaluation and were merged into one class in the penultimate column of Table 2. By comparing the MI in the merged class, the reference duos showed values of 10 and 13; the varieties with supporting data values from 1 ('Glera Lunga') to 14 ('Forgiarin'); the varieties with conflicting data values from 52 for 'Glera'/'Glera Lunga' pair to 251 for 'Marzemino'/'Marzemina Bianca' . The merged class MI was therefore at least 3.7 times higher for varieties with conflicting data than for those with supporting data. In conclusion, ASSIsT-mediated MI values better supported all the PO relationships based on PLINK parameters and still invalidated varieties with conflicting data compared to previous results reported in the literature.
Finally, Colony inferred the nSSR genotype of the presumed common parent for FS related ' Aghedene' , 'Glera' and 'Glera Lunga' varieties (P2 in Fig. 1 and Table S5). This genetic profile was checked in combination with 'Vulpea' for all three trios, and no mismatching alleles were found. The same genotype was checked for possible PO relationships with the other varieties analyzed in this study using GenAlEx and two additional perfect matches were found, one with 'Mocula' and the other with 'Pignolo' (Fig. 1).
The dendrogram of genetic similarity produced via UPGMA divided the studied genotypes into different clusters, six of them showing significant bootstrap values (Fig. 3). As can be seen, genotypes clustered together according to the parental tree shown in Fig. 1. The significant clusters, labeled with alphabetical letters and dotted lines, varied in size from five to 16 varieties. A, the family group of 'Refosco Nostrano' , 15 varieties, seven of them strictly related; this group also encompassed Teroldego, Marzemino and Refosco dal Peduncolo Rosso. B, the family group of 'Volovnik' , 'Verduzzo Friulano' and 'Picolit' , showing seven interrelated cultivars; C, the family group of 'Traminer' , with seven cultivars, five of them interrelated; D, the group of 'Heunisch' , with nine cultivars, three of them being 'Heunisch' offspring; E, the family group of 'Vulpea' , with 16 cultivars, ten of them being 'Vulpea' offspring; F, a very clearly distinct group including five cultivars related to 'Garganega' and 'Raboso Piave' as shown on the right in Fig. 1.

Discussion
Parentage analysis was undertaken in a two-step approach commonly used when searching for pedigrees in grapevine [29][30][31] . Initially, the FVG varieties genotyped with 12 nSSRs were scored for possible PO relationships using the VIVC and CREA Viticulture and Enology molecular databases, each one encompassing around 3900 unique genetic profiles. The refined dataset was then used for the analysis with the 18k SNP chip of ILLUMINA.
After filtering, 66% of the total SNP markers available in the chip were used for subsequent analyses. Our data agree with 25 that just the 14 SNPs they selected from ILLUMINA chip for identification purposes were enough to also identify all the cultivars analyzed in this study. This is interesting because SNP are qualitative markers and, unlike SSRs, do not need conversion to be comparable with the data produced in different labs using reference varieties. If further confirmed, the shared application of this set of 14 SNP markers would open the   The present FVG grapevine varietal assortment showed no cultivars derived from self-pollination. As a rule, a low number of generations was detected and some ancient varieties can be regarded as "founders", like 'Vulpea' and 'Refosco Nostrano' . Our findings agree with what was pointed out at a broader level of pedigree complexity 8 .
Surprisingly, 'Vulpea' , which was classified in the VIVC as a Romanian variety, is present in FVG, where it counted among the ancient local varieties as anonymous or even misnomered vines. For example, genotyping www.nature.com/scientificreports www.nature.com/scientificreports/ of old vines with 12 nSSRs revealed 'Vulpea' under wrong designations like 'Codelunghe' and 'Piccola Nera' 3 . Moreover, again by means of genotyping, we recognized 'Vulpea' in the nearby Veneto region, under the name of 'Quaiara' in Verona province and as 'Rossetta' or 'Sciavetta' or 'Doretta' on the Euganean Hills, near Padua, where it has been cultivated with some success to produce a popular rosé table wine. Information on 'Quaiara' and 'Rossetta/Sciavetta/Doretta' is available in 32,33 . A comparison of 'Vulpea' ampelographic description published in the European Vitis Database (http://www.eu-vitis.de, accession number ROM051-272) with 'Quaiara' and 'Rossetta/Sciavetta/Doretta' accessions grown at CREA-Viticulture and Enology repository in Susegana (TV) evidenced matching morphology (Severina Cancellier, personal communication). In the 19 th century, 'Vulpea' existed in nearly every Austrian and northeast Slovenian vineyard 34,35 and in Siebenbürgen/Romania 36 . The Austrian Helbling provided the first accurate description in 1777 under the name 'Schwarzer Abendroth' 37 . According to ampelographers of the 18 th and 19 th centuries the late maturating 'Vulpea' produced bunches with green, red and blue berries. The wine was of low quality, had a light red color and sour taste, however the cultivar was appreciated for its heavy crop. Similar to 'Heunisch Weiss' , it was recommended for "eradication" 35 . 'Vulpea' is grown successfully in the warmer Italy 34 , thus indicating that 'Vulpea' spread from neighboring countries to northeast Italy, probably since ancient times.
The parents of 'Vulpea' are the Croatian 'Bratkovina Crna' and Hungarian 'Gyoengy Feher' 8 . Therefore, according to present knowledge, all FVG cultivars first-degree related to 'Vulpea' should be offspring of this variety.
The discovery of Vulpea's high impact on the FVG variety assortment and the involvement of the Croatian 'Bratkovina Crna' and Hungarian 'Gyoengy Feher' in its ancestry pointed more to Austria as the country of origin than to Romania or Moldavia. In the latter 'Vulpea' , known as 'Ciorcuta Rosie' , was also considered autochthonous.
'Cjanorie' appears to be a bridge between 'Vulpea' and 'Refosco Nostrano' . 'Cjanorie' is PO related to 'Vulpea' and 'Vinoso Rosso' . Moreover, no pair of varieties in PO with 'Refosco Nostrano' can be the parents of this ancient variety and Colony indicates an FS relationship between 'Refosco di Rauscedo' and 'Vinoso Rosso' . Consequently, the only possible kinship for 'Cjanorie' seems to be that it is the progeny of 'Vulpea' and 'Vinoso Rosso' .
'Ribolla Gialla' cultivar of the Italian Collio corresponds to the 'Rebula' grown in Goriška Brda, the Slovenian Collio, as confirmed with genotyping 39 . The same authors analyzed a so-called 'Rebula Stara' , meaning "old Rebula". This variety was shown to be different from 'Ribolla Gialla' and possibly PO related, according to 11 SSRs. It was a surprise to discover that 'Rebula Stara' corresponded to 'Heunisch Weiss' after molecular profile comparison with available databases. 'Heunisch Weiss' is a variety of paramount importance for the evolution of grapevine varietal assortment. The pair 'Pinot' and 'Heunisch Weiss' originated a dozen French varieties 40 , among them 'Chardonnay' and 'Gamay'; in addition, 'Heunisch Weiss' could be the parent of more than a hundred varieties, even outside France, such as in Germany, Greece, Hungary and Romania 8 . Three years after the work by 39 , the hypothesis of a PO relationship between 'Ribolla Gialla' and 'Heunisch Weiss' using 20 SSRs was again supported by 8 . However, this parentage was ruled out by 9 , because five mismatching markers were found using a deep genotyping with 58 SSRs and these incompatibilities were considered too many. These authors did not take into consideration another aspect that can explain the mismatches found. In ancient varieties propagated vegetatively for centuries, molecular differences can accumulate as in a biological clock. For example, a wide clonal diversity was found in 'Savagnin' 41 , a very ancient variety at least 900 years old 42 , using only 30 SSRs: 49 plants were analyzed 41 and only 12 showed the same genotype, i.e. 24.5%. In conclusion, if the genotyping is restricted to only one plant, such differences cannot be evidenced, and molecular information could lead to erroneous conclusions. For this reason, six accessions of 'Heunisch' , coming from three countries were compared in this work, some of them being phenotypic variants 16 , and two 'Ribolla Gialla' accessions. SNP data strongly support a first-degree relationship between 'Heunisch Weiss' and 'Ribolla Gialla' . 'Robola' sounds very similar to 'Rebula' and 'Ribolla' . 'Robola' is also homonym for different varieties grown in Greece. A 'Robolla' sample coming from Greece and corresponding to 'Goustolidi' (VIVC variety number 5000) was added to our varietal list, to understand if there was some link among these varieties. All available genotyping data exclude that 'Ribolla Gialla' is currently grown in Greece as any one of the genotyped 'Robola' cultivars 9,43 . Moreover, 'Robola/Goustolidi' was shown to be another Greek variety PO related to 'Heunisch Weiss' , supporting a previous finding 8 .
Other local Italian varieties, such as 'Pinella' (Veneto) and 'Piccola Nera' (FVG) showed possible PO relationship with 'Heunisch Weiss' according to 20 SSR markers 8 ; SNP data confirmed those results. The Slovenian 'Volovnik' was shown to be first degree related to two FVG varieties, 'Forgiarin' and 'Verduzzo Friulano' , and is possibly one of their parents, while no relationship was found for 'Klarnica' .
'Mocula' and 'Pignolo' could be half-sib of ' Aghedene' , 'Glera' and 'Glera Lunga' . These additional pedigree relationships reinforce the hypothesis that the P2 vine really existed in the past. However, recovery of the P2 genotype is necessary to confirm these assumptions.
The dendrogram constructed by UPGMA is a fairly faithful representation of already known pedigree relationships and the new ones found in this research, in agreement with the relatedness proposed and summarized in Fig. 1. Most of the offspring were directly aligned to the fourteen putative parents. The dendrogram also represents a source of information to further investigate the traditional varieties genetic backgrounds or relationships.
The cluster analysis clearly separated the three most prominent founder varieties of this study 'Vulpea' , 'Refosco Nostrano' and 'Heunisch' , pointing to their distinct geographic and genetic backgrounds. These three varieties created large family groups. The fairly stringent grouping leading to the clear separation observed between variety families can be due to the conscious choice of additional cultivars (Table S2), which turned out to be related to traditional varieties in FVG. However, placement in the dendrogram can be critical in the case of "bridge" varieties, like 'Piccola Nera' ('Heunisch' × 'Vulpea') and Cjanorie ('Vulpea' × 'Refosco Nostrano'), both grouped with 'Vulpea' , or for varieties with still unclear pedigree, like 'Riesling Weiss' , grouped with 'Traminer' even if PO related to 'Heunisch' . In this last case the reason may reside in the 2 nd parent being very close to Traminer and the wild grape (Erika Maul, personal communication) and on the "attraction" by 'Manzoni Bianco' ('Pinot' × 'Riesling Weiss').

conclusions
Genotyping studies provide a more stringent and sometimes new perspective on the origin, spread and relatedness of grapevine cultivars. Research on pedigrees highlighted that outcrossing is the main strategy for the birth of new varieties, whilst very few cultivars derive from selfing; it has now become recognized that grape flowers are subjected not only to cleistogamy, but also to pollination by wind or insects. Shedding light on grapevine pedigree relationships helps in understanding the germplasm assortment evolution and cultivar history. Some varieties of FVG are shared with neighboring areas, in and outside Italy, like 'Glera' , 'Glera Lunga' , 'Ribolla Gialla' and 'Piccola Nera' , others are specific to FVG. Even their history intertwines with local and foreign varieties. Surprisingly, 'Vulpea' was shown not only to be spread in FVG and Veneto since ancient times, but it was also recognized as a recurrent parent of at least ten FVG cultivars, both ancient and well-known, like 'Glera' , as well as recent or neglected. Many 'Refosco' varieties are PO related to 'Refosco Nostrano' , showing that Refosco's represent a real family; 'Refosco dal Peduncolo Rosso' does not strictly belong to this family and is likely full-sib related to 'Marzemino' . 'Ribolla Gialla' was shown to be PO related to 'Heunisch Weiss' , therefore being another of the numerous offspring derived from this prolific variety.
In conclusion, combining molecular markers and historical references was shown to be a high-performance strategy to retrace and adjust the history of cultivars.