Introduction

Italy is among the most established and top-quality wine-producing countries in Europe, with viticulture having increasingly become one of the leading businesses of the agri-food economy. Furthermore, Italy has one of the widest ampelographic platforms, constituting a wealth of grapevine varieties that can be considered an invaluable source of plant biodiversity (Bacilieri et al., 2013; De Lorenzis et al., 2019).

In the last 150 years, the Italian total vineyard surface area has undergone a sharp reduction, especially in some agricultural districts, decreasing from over one million hectares in the 1970s to approximately 700,000 hectares today (Boselli et al., 2019; OIV - International Organisation of Vine and Wine., 2022). There are several concurrent causes of this drastic abandonment of vine cultivation (Torquati et al., 2015): i) industrialisation and urbanisation (above all, in certain regions such as Lombardy, one of the most productive of northern Italy) (Raimondi et al., 2015), ii) rural depopulation (a chronic outmigration of mostly young adults looking for job opportunities elsewhere), iii) a significant increase in the pressure of vine pests of American origin, predominantly phylloxera, downy mildew and powdery mildew (Fontaine et al., 2021; Gadoury et al., 2012; Granett et al., 2001). Moreover, many traditional autochthonous cultivars have been replaced by the so-called “international varieties” (i.e., Cabernet Sauvignon, Merlot, Sauvignon Blanc, Chardonnay), comprising a few genotypes that determine most of the worldwide grapevine diversity (Wolkovich et al., 2018), especially due to their high adaptability to different environments (Otto et al., 2022).

This sudden evolution of the grapevine landscape has caused genetic erosion, both in cultivated and wild grapevines (Gisbert et al., 2018). Thus, several varieties once commonly grown and cited in the most popular ampelography treatises are no longer available or are in danger of disappearing. For these reasons, many research groups working in the regions traditionally suited to viticulture are engaged in the protection of ancient autochthonous, less known or forgotten vines (Ghrissi et al., 2022; Goryslavets et al., 2015; Gristina et al., 2017; Maraš et al., 2014; Margaryan et al., 2021; Moravcova et al., 2006; Sargolzaei et al., 2021; Schneider et al., 2014; Zombardo et al., 2021).

The wine-growing area involved in this study is located in Tuscany (central Italy), one of the most renowned regions in the world from an agricultural, touristic and landscape point of view. In particular, Candia dei Colli Apuani, Garfagnana, and Lunigiana are three wide valleys in the northwest coastal area, on the border of Levante Ligure (east Liguria), lying between the Apennines and the Apuan Alps and overlooking the Tyrrhenian Sea. Historically, these territories maintained broad administrative independence and have been relevant for maritime trade with foreign countries or land trade with the rest of Italy (Piccardi and Pranzini, 2018). As already mentioned, the vineyards also in these places are disappearing. The main reason is land abandonment with a subsequent advanced average age of farmers, exacerbated by environmental constraints (i.e., harsh sites, significant elevations, steep slopes, hydrogeological hazards) and the high pressure of fungal diseases linked to abundant rainfall (over 1200 mm/year) (García-Ruiz and Lana-Renault, 2011; Jackson, 2008; Storchi et al., 2018). In this area, so-called “extreme viticulture” is still carried out, consisting of terraced vineyards confined to isolated areas with limited accessibility and thus associated with the impossibility of mechanised agronomic practices and prohibitive management costs (Strub and Loose, 2021). Consequently, there are no major commercial wineries but only family-run production units, 90 % of which have a fragmented cultivated area, often smaller than 5 hectares (Storchi et al., 2018).

The growing concern about land abandonment and the low profitability of vineyard production has resulted in actions being carried out aiming to recover old vine varieties. Native grapevine germplasm renewal could prove fundamental in not only avoiding genetic erosion but also achieving differentiated productions and better adapting plant material to the local climatic conditions. Considering the viticultural vocation of this important transit area, it is likely that a rich biodiversity can be found. The heterogeneity of the grapevine germplasm in remote districts is, in fact, an indicator of the movement of people and goods, and it is assumed that the ampelographic heritage has been enriched by grape varieties imported from elsewhere over the centuries.

The research work described here aimed at rescuing individual vines discovered along field borders or in areas that are no longer cultivated and are located in different municipalities of the Candia dei Colli Apuani, Garfagnana, Lunigiana and Levante Ligure territories. Genotyping of 99 recovered vines was carried out using a set of 12 SSR markers (Simple Sequence Repeats), an internationally recognised and widely used method for grapevine genetic identification (Pelsy, 2010). The rarest and most interesting genotypes were propagated and gathered in a grapevine germplasm collection to allow ex-situ conservation and phenotypic characterisation.

Materials and methods

1. Plant materials

The plant materials (young leaves or small shoot portions) of 99 relic vines were sampled from 15 different Italian municipalities belonging to the territories of Candia dei Colli Apuani (2, Massa Carrara province), Garfagnana (3, Lucca province), Lunigiana (8, Massa Carrara province) and Levante Ligure (2, Genoa province), encompassing an area of approximately 1050 km² (Figure 1).

The vine specimens (all non-grafted) were identified in reforested wine-growing areas or field borders by local farmers who carried out visual inspections. After the plant material collection, each sample was named after the original location, the contact person or the putative varietal name referred to (Table 1). The sampling took place progressively, starting from the early 2000s and based on reports relating to viticultural material found within the wide area of interest. Confirmatory genetic analyses were carried out on all samples in 2021 and 2022.

Figure 1. Geographical locations and list of the 15 Italian municipalities from which the 99 grapevine samples included in the study were taken.

2. SSR Genotyping

Genotyping was performed using 12 SSR markers: nine as the minimum standard set for grapevine identification (VVS2, VVMD5, VVMD7, VVMD25, VVMD27, VVMD28, VVMD32, VrZAG62, VrZAG79) (Maul et al., 2012), plus VMC6E1 (ISV2), VMC6G1 (ISV4) (Crespan, 2003) and VMCNG4b9 (Welter et al., 2007). PCR reactions were performed using forward primers labelled with fluorescent dyes (6-FAM, PET, VIC, or NED); two multiplex panels of fluorescent-labelled microsatellite loci were used. Simultaneous PCR amplifications were conducted in a final volume of 12.5 μl containing 1× PCR reaction buffer, 10 ng of genomic DNA, 0.2 mM of each dNTPs, 2 mM MgCl₂, 1.5 U Taq DNA Polymerase (Thermo Fischer Scientific, Waltham, MA). Depending on the locus, primer concentrations ranged from 0.11 to 0.48 μM (see Additional file 1). Reactions were performed on a GeneAmp PCR System 9700 using the following profile: a hot start of 95 °C for 5 min, 30 amplification cycles of 45 sec at 95 °C, 1 min at 55 °C, 30 sec at 72 °C, and a final extension step of 30 min at 72 °C. 0.5 µL PCR product for each sample was mixed with 9.35 µL of formamide and 0.15 µL of GeneScan™ 500 LIZ Size Standard (Life Tech, Carlsbad, CA, USA). Afterwards, the capillary electrophoresis was run using an ABI 3130xl Genetic Analyzer (Life Tech, CA, USA). Allele calling was performed employing GeneMapper version 5.0 (Thermo Fisher Scientific, Waltham, MA, USA) and using a homemade bin-set built with some reference varieties. Allele sizes were recorded in bp (using the VIVC allele sizing), and genotypes showing a single peak at a given locus were considered homozygous. The resulting SSR profiles were compared with those in the CREA Viticulture and Enology molecular database (a constantly updated repository that to date contains 4982 unique Vitis spp. microsatellite profiles), in the Vitis International Variety Catalogue (https://www.vivc.de/), and in the literature to ensure unambiguous varietal identification.

3. Statistics on SSR Markers

Statistics on the SSR markers data were computed using Cervus 3.0.7 software (freely available at http://www.fieldgenetics.com/pages/home.jsp) and GenAlEx 6.51b2 software (Peakall and Smouse, 2012). Recognised hybrids (6 samples) were excluded from the computation. Cervus was also used to infer putative trios (parent-parent-offspring related genotypes). The search for parent-offspring relationships, however, was limited to the 58 genetic profiles found in this work.

4. Genetic Similarity and Ancestry Evaluation

Genetic distances were computed with GenAlEx 6.51b2 software (Peakall and Smouse, 2012). Recognised hybrids were excluded from computation. PCoA was also performed using GenAlEx software. Genetic distances were used to run MEGA X - Molecular Evolutionary Genetics Analysis across computing platforms (Kumar et al., 2018) software, version 10.0.5 for a NJ (Neighbor-Joining) phylogenetic tree elaboration (Saitou and Nei, 1987).

Ancestry values and number of subpopulations were calculated on 52 genotypes, excluding hybrids, using STRUCTURE 2.3.1 software (Pritchard et al., 2000). Ten independent runs were performed for K values from 1 to 8, with a burn-in period of 50,000, followed by 50,000 MCMC replications. The best K was calculated using the ΔK parameter (Evanno et al., 2005). The chosen membership coefficient threshold for membership to a given cluster was Q = 0.80. Data reported in Additional file 2 were obtained after an alignment of the 10 independent runs calculated from K = 2 to K = 8 using CLUMPP software (Jakobsson and Rosenberg, 2007), version 1.1.2 (freely available at https://rosenberglab.stanford.edu/clumpp).

5. Grapevine Germplasm Collection

From the year 2005 onwards, an experimental vineyard with the purpose of local viticultural germplasm collection was set up at “Podere Scurtarola” farm in the municipality of Massa (Tuscany, Italy; 44°04’80” N, 10°10’98” E, 200 m a.s.l.), covering an area of approximately 0.5 hectares. In this environment, the average annual rainfall is about 1200 mm (data from Consorzio LaMMa, available at https://www.lamma.rete.toscana.it/), and a mean Huglin index value of 2010 °C (Santos et al., 2012).

The vineyard is terraced and situated on a steep slope (35°, corresponding to a gradient of more than 70 %). The soil has a sandy clay loam texture (63 % sand, 25 % clay, 12 % silt), and the inter-row has permanent spontaneous grass cover to reduce erosion and the risk of landslides.

A preliminary virus screening was carried out on additional plant material from the sampled vines, according to the directives of the Italian clonal selection protocol for grapevine (Grapevine fanleaf virus – GFLV; Grapevine leaf roll associated virus 1, 2, and 3 - GLRaV-1,-2, 3; Grapevine virus A - GVA, Grapevine virus B - GVB; Grapevine fleck virus - GFkV; Arabis mosaic virus - ArMV). Portions of woody branches taken from the mother plants of the formerly screened virus-free samples were propagated by grafting on 1103 Paulsen rootstock. The root cuttings (at least 20 specimens each) were planted in the vineyard with an east-west orientation and a 0.8 x 2.30 m row spacing. Vermentino (VCR1 clone) was added as a reference variety, being the main grape cultivar in the area of interest. The vines were trained on an upward vertical shoot-positioned trellis, with spur cordon pruning and an average of 10 buds per vine. Over the years, homogeneous agronomic conditions were maintained on the vineyard and organic pest management was carried out.

6. Ampelographic Descriptions

The ampelographic description of some of the main vine organs was carried out on the 47 accessions of already adult vines (over 3 years old) grown in the vineyard collection; these are reported in bold and an asterisk in Table 1. The ampelographic characteristics were recorded according to the recommended OIV - International Organization of Vine and Wine methodology (descriptor list for grape varieties and Vitis species, available at https://www.oiv.int/public/medias/2274/code-2e-edition-finale.pdf), both for the unknown and the already known genotypes. A set of 14 standardised OIV primary descriptors (priority list) was used: young shoot (OIV 001, 004), shoot (OIV 016), young leaf (OIV 051), mature leaf (OIV 067, 068, 070, 076, 079, 081-2, 084, 087) and berry (OIV 223, 225). Moreover, the OIV 151 descriptor for flower sexual organs was added.

7. Anthocyanin Content and Pigment Profiles of Berry Skins

The anthocyanin content and pigment profile of berry skins for seven coloured-berry unknown genotypes were determined following the method described by Gomez-Alonso (Gómez-Alonso et al., 2007) with some modifications. Instead of separating the skin from the pulp, 100 g aliquots of entire berries were blended at maximum speed for 45 s using a 6805 model Osterizer blender (Sunbeam, FL, USA). The pigments were then extracted from 5 g of the resulting blend by adding 5 mL of a 3 % formic acid solution (CH₂O₂) to methanol (CH₃OH) (HPLC grade, Carlo Erba, Italy). After 45 min on an orbital shaker, the samples were centrifuged at 4000 rpm for 5 min, and the supernatant was filtered through a 0.45 µm RC syringe filter (Cole-Parmer Instrument Co. Europe, United Kingdom) and collected in 2.5 mL glass vials with screw caps. HPLC analyses were performed by injecting 10 µL of the sample into an HPLC system (Agilent Technologies, CA, USA) composed of an autosampler (1260 series), solvent degasser (1100 series), quaternary pump (1100 series), thermostatted column compartment (1100 series) and diode array detector (DAD, 1200 series). HPLC was driven by a Personal Computer running Agilent ChemStation for LC 3D System software (Agilent, California, USA). Anthocyanins were separated using a Luna® Omega 5 µm Polar C18 100 Å 250 x 4.6 mm column (Phenomenex, CA, USA) preceded by a guard column packed with the same material, maintained at 10° C for better separation of the pigments. Three different mobile phases were employed: mobile phase A comprised a 50 mM ammonium dihydrogen phosphate solution (NH₄H₂PO₄) in water (H₂O) with pH adjusted to 2.60 with 85 % phosphoric acid (H₃PO₄) (analysis grade, PanReac, Spain); mobile phase B comprised 20 % v/v of the mobile phase A in acetonitrile (C₂H₃N) (Carlo Erba, Italy); and mobile phase C comprised a 200 mM phosphoric acid solution (H₃PO₄) in water (H₂O) at pH = 1.5. The mobile phase gradient is described in (Gómez-Alonso et al., 2007); the chromatograms were collected at 520 nm, and the pigments were identified using peak retention times suggested by the authors. Pure standards of malvidin-3-O-glucoside chloride (C₂₉H₃₅O₁₇Cl – Extrasynthèse, France) were injected to establish a calibration curve for the quantitative results.

Results

1. Genetic Identification of Grapevine Samples

The molecular analyses performed on the collected plant materials allowed us to identify 79 vines, while 20 remained anonymous (Table 1): 73 samples corresponded to 36 already known Vitis vinifera varieties, 6 samples to 6 Vitis interspecific crosses, and 20 samples to 16 hitherto unreported genotypes. The 58 different SSR profiles obtained are shown in Table 2. Only Unknown 04, Unknown 08, Unknown 10 and Unknown 11 were found twice. The remaining unknown genotypes could be actual cultivars in high danger of extinction or plants obtained from single seeds. Unknown 03 is likely a hybrid (never genotyped as a rootstock), because it displays some alleles that are absent in V. vinifera, namely allele 237 at VVMD7, allele 133 at VMC6E1 (ISV2) locus and allele 179 at VrZAG62 (Crespan et al., 2009). This observation was also validated through a comparison with the up-to-date CREA-VE molecular database.

2. Statistics on SSR data and trio relationships

Diversity parameters were computed for the 12 detected SSR markers on 52 genotypes (hybrids were discarded). The results are reported in Table 3. The number of alleles found per locus among the samples ranged from 6 for VVMD5 and VMC6G1 (ISV4) loci to 10 for the VVMD28, VrZAG79, VMC6E1 (ISV2) and VMCNG4b9 loci, with a mean number of 8.67. The number of effective alleles is 4.896, and Shannon’s index is 1.774. The mean observed heterozygosity is higher than expected (0.838 vs 0.789), as supported by the fixation index, negative for all loci and suggesting no inbreeding. The PIC values ranged from 0.667 for the VVMD25 locus to 0.806 for VMC6E1 (ISV2), with an average of 0.762. Overall, the combined non-exclusion probability (identity) value was very low: 5.59E-09. The probability of null alleles was negative for all loci.

Two trio combinations were found, in which the putative offspring share at least one allele per locus with its possible parents: Gallizzone may originate from a spontaneous cross between Farinella and Sciaccarello and Unknown 10 from Durella Gentile and Pollera Nera (Table 4). Given the low number of molecular markers used for this kind of comparison, this preliminary information needs to be reinforced with additional molecular data.

3. Genetic Similarity and Ancestry evaluation

The genetic distances computed with GenAlEx software are represented by the Principal Coordinate Analysis displayed in Figure 2. The first axis explains 9.76 % of the variation, while the second explains 7.97 % (for a total of 17.73 %). Unknown genotypes are evenly distributed in three of the plain quadrants.

Figure 2. PCoA results obtained by comparing the genetic distances of 52 genotypes.

The NJ tree inferred with the genetic distances (12 SSR markers from 52 genotypes) is shown in Figure 3. Even when representing the data in this way, the unknown genotypes are scattered throughout almost every branch of the tree. Interestingly, the Unknown 12 genotype can be seen to be very different from all the others and can be considered an out-group together with the Cabernet Sauvignon and Ancellotta cultivars.

Figure 3. NJ tree of 52 genotypes with 12 SSR markers.

When considering ancestry evaluation, the search for the best number of subpopulations (K) resulted in the most significant value being found for K = 2 (Figure 4). Seventeen genotypes stand out from the rest of the population, being or having been among the grapevine varieties typical of Liguria and Tuscany, like Albarola, Caloria, Ciliegiolo, Della Borra, Durella Gentile, Farinella, Gallizzone, Pollera Nera and Sciaccarello, plus eight of the unknown genotypes (Additional file 2). Twenty-three genotypes were grouped in the second cluster and encompassed very different varieties, from the ancient table grape Afus Ali (belonging to proles orientalis) to the wine grape Cabernet Sauvignon (belonging to proles occidentalis) (Negrul, 1946); this second group also comprised five unknown genotypes, while twelve genotypes were admixed. Observing the groups of genotypes with Q > 0.80 for increasing K values, a nucleus of four genotypes remained fixed up to K = 8; these were Caloria, Pollera Nera, Sciaccarello and Unknown 14.

Figure 4. Estimate of the number of clusters obtained with STRUCTURE for K values ranging from 1 to 8 using 12 SSR.

Overall, a confluence between the PCoA (Figure 2) and STRUCTURE results was noted. Interestingly, the group of 17 genotypes found with STRUCTURE analysis for K = 2 occupies the left side of the PCoA plain, even if not in an exclusive way. Moreover, several unknown genotypes enlarge the plain of genetic distances much more than renowned varieties to be different from each other, like Afus Ali and Cabernet Sauvignon; two additional unknown genotypes are positioned on the left-hand side, opposite Afus Ali and Cabernet Sauvignon.

4. Ampelographic Descriptions within the Grapevine Vineyard Collection

As a result of the profitable activity of local grapevine germplasm recovery implemented over the years, the vineyard collection has come to host more than half of the samples discovered (52 out of 99) following vegetative multiplication and planting. Currently, the vineyard contains 39 known and 12 unknown grapevine accessions, plus one Vitis interspecific cross. Among these, 47 accessions have over-3-year-old vines: 36 known grapevine varieties belonging to 18 different genotypes, 1 Vitis interspecific cross (Jacquez), and 10 belonging to 8 unknown genotypes, including a putative hybrid, Unknown 03 (Table 1). The ampelographic characteristics of adult vines were assessed and are shown in Table 5, limited to the 8 unknown genotypes. As for the 19 known genotypes, we verified that the morphology of the main organs correctly corresponded to the ampelographic descriptions already present in the official databases (data not shown, available on request).

5. Phenotypic description of the grapes for unknown genotypes: anthocyanin contents and pigment profiles

Seven unknown genotypes with coloured berries were characterised at the phenotypic level by analysing the quantity and quality of the skin anthocyanins (see Table 6).

The anthocyanin content was highly variable and reflected the visual characteristics detected on the bunch (Table 5, OIV descriptor 225). The genotypes with grey and pink berries (Unknown 07 and Unknown 15) exhibited only small amounts of anthocyanins, whereas quantities 20 times higher were found in Unknown 10 and Unknown 11.

Malvidin-3-O-glucoside was the most prevalent anthocyanin in Unknown 11 and Unknown 16, accounting for 45 % and 50 % respectively of total anthocyanins. Furthermore, these two genotypes shared a similar profile; in both cases, trisubstituted anthocyanins made up approximately 50 % of the total, and the proportion of acylated anthocyanins was around 36 %, with malvidin-3-O-p-coumaroylglucoside representing more than 20 % of total anthocyanins. Unknown 07 also had a higher percentage of trisubstituted anthocyanins, but with a more homogeneous distribution among the different pigments. The most represented were delphinidin-3-O-glucoside, cyanidin-3-O-glucoside and malvidin-3-O-glucoside, each with values above 20 %. Approximately 14 % of acylated anthocyanins were almost entirely represented by malvidin-3-O-acetylglucoside, delphinidin-3-O-p-coumaroylglucoside and malvidin-3-O-p-coumaroylglucoside. Unknown 08, 10 and 15 were characterised by a low percentage of trisubstituted anthocyanins. In Unknown 15, cyanidin-3-O-glucoside represented almost 70 % of the pigments, and, except for malvidin-3-O-glucoside (11.50 %), no other anthocyanin reached 5 %. Peonidin-3-O-glucoside represented approximately 60 % of the anthocyanins of Unknown 10, followed by malvidin-3-O-glucoside (22 %) and cyanidin-3-O-glucoside (7.4 %), while acylates were the lowest among all the samples (3.6 %). Unknown 08 was the genotype with the lowest percentage of trisubstituted anthocyanins, owing to the high percentages of cyanidin-3-O-glucoside and peonidin-3-O-acetylglucoside, which together accounted for almost 80 % of all anthocyanins. Finally, the most abundant anthocyanins in Unknown 13 were malvidin-3-O-glucoside and peonidin-3-O-glucoside, which, with similar percentages, constituted approximately 70 % of the profile. The percentages of trisubstituted and disubstituted anthocyanins were similar, while acylated anthocyanins represented about 10 % of the total pigments.

Discussion

1. Research background, safeguard of viticultural germplasm reservoirs and biodiversity genetics

Italian viticulture is characterised by a surprising richness of grapevine varieties, the legacy of centuries of natural selection and human cultivation, considering that the winegrowing tradition in this country dates back to the Protohistoric Age (Crespan et al., 2023; Mannini, 2004). Rescuing forgotten grapevines and characterising retrieved unknown genotypes can contribute to providing a valuable source of viticultural biodiversity in the face of climate change (Van Leeuwen and Destrac-Irvine, 2017), as well as putative missing links in kinship studies of the noblest vines (Raimondi et al., 2020) and a cultural and productive heritage for their place of origin (Gutiérrez-Gamboa et al., 2020; Mannini, 2004). Beyond studies that have already been performed on grapevine germplasm collections and that give a general idea of existing biodiversity (Cipriani et al., 2010; Cretazzo et al., 2022; Emanuelli et al., 2013; Laucou et al., 2011; Maraš et al., 2014), there are several recent or ongoing studies in Italy based on SSR genotyping related to the in situ discovery of niche varieties that have survived in circumscribed remote environments (Fanelli et al., 2021; Pastore et al., 2020; Zombardo et al., 2021).

The coastal area of Tuscany (less famous than other wine districts in this region, such as Chianti and Montalcino), where this research project was implemented, has a strong tradition in viticulture that has undergone a sharp decline due to both environmental and anthropogenic factors (Schultz and Jones, 2010; Storchi et al., 2018). The study carried out in old, abandoned vineyard plots scattered across the Candia dei Colli Apuani, Garfagnana, Lunigiana, and Levante Ligure territories (Figure 1) allowed us to correctly identify the 99 grapevine samples collected over the years, obtaining an approximate idea of the past varietal composition. In detail, we highlighted some cases of synonymy between official and local names, found mismatches with the putative varietal name reported for some samples and identified 16 hitherto unreported genotypes (Table 1).

Interesting information emerged from the in-depth analysis of the obtained molecular results. Primarily, the number of alleles per locus revealed a vast biodiversity within the recovered pool of samples (Table 3). Even the identified trio combinations provide intriguing insights. When considering the possible origin of Gallizzone as a spontaneous cross between Farinella and Sciaccarello (Table 4), it should be noted that the latter variety was deemed one of the main parent varieties within the assortment of Italian grapevine germplasm, especially regarding central Italy (Crespan et al., 2023). Therefore, this preliminary indication is well supported by available information, at least on one of the putative parent candidates.

An ancestry evaluation was carried out using available data. Considering that the Italian grapevine germplasm was found to be admixed, until a clear distinction emerged from the different evolutionary history between Northern and Southern Italian grapevines (Mercati et al., 2021), and that the search for subgroups is usually applied to hundreds of genotypes (Bacilieri et al., 2013; Cipriani et al., 2010; Cretazzo et al., 2022), we nevertheless obtained interesting results from our limited population. The split of our pool of samples into two sub-populations allowed us to highlight one first group of 17 genotypes, of which eight (Caloria, Della Borra, Gallizzone, Pollera Nera, Sciaccarello, and three unknown genotypes) remained constantly joined, even for increasingly clustered groups (up to 7), until a complete population destructured for K = 8. It is worth noting that Della Borra, Pollera Nera, and Caloria have been recognised as progenies of the barely mentioned Sciaccarello (D’Onofrio et al., 2021). Given the restricted geographical area of sample collection, we can assume these genotypes to be the main grapevine varieties characterising the study area.

2. Contextualised information about known genotypes

Of the known grapevine varieties retrieved in the sample pool, the most represented genotypes, with at least five individuals each, were Vermentino (as expected), Agostenga, Barsaglina and Ciliegiolo. Vermentino is the most cultivated variety in this geographical area, and the white easy-to-drink wines produced from its grapes on the Tuscan coast are highly appreciated. For this reason, it was included in the vineyard collection as a reference variety. Agostenga is an early-ripening, white-berried variety grown today exclusively in some high mountain areas of Aosta Valley (where it is known as Prié Blanc or Blanc de Morgex). Recently, a synonymy with the Spanish cultivar Legiruela was confirmed, suggesting that it was moved to very geographically distant areas several centuries ago (Schneider et al., 2010). All five samples were taken from an area of mid-elevation in the province of Lucca (Minucciano, about 700 a.s.l.). In Italy, Agostenga has been identified among relic vines in Garfagnana in previous studies (D’Onofrio et al., 2016), as well as in Southern Umbria (Zombardo et al., 2021), suggesting that in Central Italy, while no longer in existence, there was a minimal presence of this cultivar in the past. This sporadic presence may be due to the introduction of Agostenga as a grape for fresh consumption, as suggested by the reference to the month of August in its name.

Barsaglina is a minor, black-berried variety from Massa Carrara, hence the official synonym Massaretta (Robinson et al., 2013). This variety is cultivated in an area of less than 20 hectares located between Tuscany and Liguria, according to the last Italian census (2010); however, there are still concrete traces of a past diffusion of this variety in the territories considered.

Ciliegiolo (Italian for “small cherry”, referring to the cherry aroma of its grapes), already cited as being one of the varieties grown around Florence in the 1600s (Soderini, 1590) is nowadays quite widespread in many Italian regions. It is now known that Ciliegiolo descends from Sangiovese and Muscat Rouge de Madère (or Moscato Violetto) (Cipriani et al., 2010; D’Onofrio et al., 2021).

Two more represented grapevine varieties in the pool of collected samples are Albarola and Bonamico. The first is a white-berried, late-ripening variety typical of “Cinque Terre” (Torello Marinoni et al., 2009). The second is a high-yield, late-ripening black variety spread all over Tuscany in the past, but now just hanging on in old vineyards in the Pisa, Pistoia and Lucca provinces (Robinson et al., 2013). Three out of the four assigned sample names were incorrect, as both Colombana and Groppello are different varieties, genetically distant from Bonamico.

Three samples collected in the municipality of Pontremoli (MS) were called “Morone” (Table 1); however, our DNA analyses revealed Morone 2 and Morone 3 to be Parmesana, and Morone 1 (with 2 other samples from Garfagnana) to belong to the Uva Crova genotype. Parmesana is a minor vine of the Emilia Romagna region (central Italy), which can be found under the name of Brugnera or Covretto in the area of Reggio Emilia (Pastore et al., 2020; Raimondi et al., 2015), Cavazza in Modena and Pallona/Giuncanesa in the bordering Garfagnana (D’Onofrio et al., 2016), all regions under the aegis of the Este Duchy in the past. Uva Crova is a vine retrieved in Candia dei Colli Apuani, historically found in Garfagnana under the name of Carraresa (D’Onofrio et al., 2016; Roncaglia, 1850) and in the Reggio Emilia province as Covra (Pastore et al., 2020). Despite the two different genetic profiles, these locally named Morone samples are all included in the “Crova, Covra, Covretto, Croa, and Croatina” family, given that the grapes have some morphological characteristics in common (Rossoni et al., 2001).

Three samples, all found in Lunigiana, belonged to the Pollera Nera genotype, one of the most representative cultivars of this area, where it is historically present (Acerbi, 1825). A recent kinship analysis revealed Pollera Nera to descend from the parent Sciaccarello, which was retrieved in this work along with some of its proven offspring: Caloria (which is also known under the generic synonym of Pollera, but has a different genetic profile from Pollera Nera), Della Borra (also locally known as Bracciola Nera, but genetically distinct from the “official” Bracciola Nera accession 1639 on VIVC), and Rollo (official synonym of Livornese Bianca in Tuscany, or Bruciapagliaio in Liguria) (Crespan et al., 2021; D’Onofrio et al., 2021; Di Vecchi Staraz et al., 2007; Torello Marinoni et al., 2009).

Two samples corresponding to Muscat of Hamburg were also found. This fine black table grape variety (progeny of Schiava Grossa x Muscat of Alexandria) (Crespan, 2003) is historically cultivated in Tuscany (mainly on the Lucca hills), where it is still used as a wine grape. This interesting example highlights how varieties of exogenous origin introduced to the local assortment become part of the production of niche wines that distinguish and enhance a small wine-growing region.

Vermentino Nero is a minor traditional vine of the province of Massa Carrara (Roncaglia, 1850) which is nearly extinct. It is nowadays grown within a few hundred hectares limited to this area and in Lunigiana, where it is also known as Merla. Vermentino Nero has a low degree of kinship with the much better-known white Vermentino; however, recent results indicate Vermentino Nero has a close genetic link with Canaiolo Nero, assuming that these two varieties share one still unknown parent (Raimondi et al., 2020).

The finding of six different Vitis interspecific crosses (five proven; 1 putative: Unknown 03) indicates that there was a sizeable cultivation of direct-producer hybrids in the past, even in central Italy (Zombardo et al., 2021). During the last century, hybrid vines were employed for stemming fungal diseases of foreign origin, but their use was gradually abandoned and even banned in the European Union in 1979 (Kapusta et al., 2018).

3. Unknown genotypes: genetic and phenotypic characteristics

The proportion of unknown grapevine varieties sourced within the pool of samples was quite significant (20 out of 99), which is more than a fifth of the total amount. This demonstrates that grapevines that do not conincide with any genotype included in the official databases had some relevance in the early 20th century. Some samples had given names that turned out to be incorrect, not corresponding with the putative variety (e.g., Schiava Bianca and Vite Mammolo). And for some others valid matches were found by comparing the ampelographic descriptions with the literature of the past (e.g., Tané, and Rossara/Monferrato) (Roncaglia, 1850).

In addition to the ampelographic description of the main vine organs (using the set of 14 standardised OIV primary descriptors), we chose to examine the anthocyanin profile of berry skins in our preliminary phenotypic evaluation of unknown genotypes. This parameter represents a relatively stable quality trait that is an expression of some variety-dependent genes (Massonnet et al., 2017) and can be used for chemotaxonomic classifications of red-berried Vitis Vinifera cultivars (Mattivi et al., 2006). In red wine grapes, higher percentages of trisubstituted anthocyanins (delphinidin-3-O-glucoside, petunidin-3-O-glucoside, and malvidin-3-O-glucoside), as well as a high proportion of acylated form of anthocyanins (acetyl-glucosides, caffeoyl-glucosides, and p-coumaroyl-glucosides) are desirable, since they enhance colour stability and confer a better aging capacity to wines (Cheynier et al., 2006; Wang et al., 2023; Zhao et al., 2017).

The results (Table 6) indicate that the most valuable genotype for red winemaking is Unknown 11, due to its high percentage of malvidin-3-O-glucoside and acylated pigments. Unknown 16 has an equally interesting anthocyanin profile, despite its low anthocyanin content (mg/kg). Such composition can be crucial in conferring good stability to the colour matrix, enabling the production of wines with a vibrant pink hue.

Conclusions

The results of this research have shed some light on the composition of the autochthonous grapevine germplasm of Candia dei Colli Apuani, Garfagnana, Lunigiana and, to a lesser extent, Levante Ligure. It was possible to identify and characterise numerous known and unknown genotypes and to create a valuable vineyard collection that includes most of them. The exploitation of the precious grapevine reservoir provided by these wine-growing regions has ensured the protection of emblematic cultivars thought to be extinct or to have almost disappeared. Our work has laid down solid foundations for the potential use of certain grapevine varieties that were probably once preferred and can henceforth be cultivated again to help revive the eco-wine tourism of the Tuscan heroic vineyards.

Acknowledgements

The authors thank the colleagues of CREA - Research Centre for Viticulture and Enology of Arezzo for their help in field surveys, plant material samplings and laboratory analyses. This research was also supported by the Service for grapevine identification of CREA - Research Centre for Viticulture and Enology of Conegliano (TV).