Chemical and sensory properties of South Tyrol red wines from disease-resistant and Vitis vinifera cultivars

Abstract

Disease-resistant hybrid grape cultivars (DRHGCs) are hybrids of Vitis vinifera and other Vitis species that provide greater resistance to specific diseases, including downy and powdery mildew. These have many advantages, but often have unusual chemical and sensory properties. This study examined the link between the chemical and sensory properties of DRHGC red wines from South Tyrol, Northern Italy, using ‘modified rate-all-that-apply’ sensory analysis, projective mapping (‘napping’), and analytical chemistry. The V. vinifera wine was more astringent, with more strawberry and jam aromas, than the DRHGC wines; the DRHGC wines had stronger green bell pepper flavours and aromas. The polyphenol and volatile profiles of DRHGCs were also distinctive, as were the profile of macrocyclic (crown) oligomeric proanthocyanidins. Panellists rated the DRHGC wines as high quality, suggesting that V. vinifera wines are not inevitably preferred.

Introduction

Sustainability has become a critical issue for agriculture in general and for wine production in particular. Viticulture uses a disproportionate amount of phytosanitary treatments because Vitis vinifera L. is highly susceptible to disease^1,2,3. There is considerable political and social pressure to reduce the usage of phytosanitary treatments across all agricultural sectors, including the European Green Deal and the EU “Farm to Fork Strategy” within the EU^2,4,5 and the Action Plan to Realize Zero Growth in Pesticide Use by 2020 in China⁶. At the same time, climate change is projected to bring increased temperatures to most viticultural regions, increasing disease pressure⁷.

It is suggested that disease-resistant hybrid grape cultivars (DRHGCs) can solve both problems since DRHGCs have been bred for disease resistance or disease tolerance. DRHGCs therefore do not require as many treatments with phytosanitary products as V. vinifera. This may also reduce the usage of fossil fuels and labour use, since fewer passes through vineyards using tractors are required, and improve worker health and safety by reducing potential exposure to harmful agrochemicals².

It is worth noting, though, that DRHGCs differ in the specific diseases they are resistant to and their level of resistance. Most commonly, DRHGCs are bred to resist downy mildew (Plasmopara viticola (Berk. & M.A. Curtis) Berl. & De Toni) and powdery mildew (Erysiphe necator Schwein.)⁵. A smaller number of DRHGCs have been bred for resistance to grapevine phylloxera (Daktulosphaira vitifoliae (Fitch 1855); primarily rootstock cultivars), Esca⁸, black rot (Phyllosticta ampelicida (Engelm.) Aa)⁹, and Pierce’s disease (Xylella fastidiosa Wells et al. 1987)¹⁰.

In addition, the chemical composition of DRHGCs is distinct from that of V. vinifera cultivars and poses challenges for the winemaking process. This is an issue that is still being addressed, with newer cultivars selected to minimise such negative characteristics such as herbaceous and ‘foxy’ aromas². Current literature was reviewed recently by Duley et al.². Briefly, DRHGCs are often high in titratable acidity (particularly malic acid), but relatively high in pH due to the higher potassium levels in the grapes. They are also often low in total soluble solids and tannins, but high in yeast-available nitrogen, proteins, and polysaccharides. The latter two in particular can lead to reduced astringency in the wine. Red DRHGCs also have a particularly distinctive anthocyanin profile, due to the presence of diglucoside and potentially triglucoside anthocyanins^2,11,12.

These differences in chemical profile can in turn can lead to a unique flavour and aroma profile that consumers may find challenging. However, little work has been done to link the sensory profile and wine chemistry of these wines. The most concerning issue is the presence of ‘foxy’ and herbaceous aromas in some DRHGCs (especially older cultivars). ‘Foxy’ aromas are caused by methyl anthranilate and 2-aminoacetophenone and are a particular issue in cultivars with genes from V. labrusca L., while herbaceous aromas are caused by eugenol, cis-3-hexenol, and methoxypyrazines².

Due to the unfamiliar flavour profile, one difficulty for DRHGC wine producers has always been a lack of market acceptance. In addition, DRHGC wines will not bear the familiar names that consumers (and therefore producers) in many countries look for^2,13,14. However, this may be changing: anecdotally, wineries in parts of the eastern USA have reported that their DRHGC wines are popular and sell well to consumers, but that members of the wine trade are resistant because of their perception of consumer resistance¹⁵. This has proven true in the Finger Lakes AVA of New York state, USA, where DRHGCs remain popular even as V. vinifera is increasingly planted. This region has an extreme climate and was initially planted solely with cold-hardy DRHGCs and V. labrusca cultivars. Here, DRHGCs can be planted in colder sites that lack the moderating climatic influence of the lakes and the lower production costs can allow for the production of less expensive wines¹⁶. In addition, it has been shown that consumer resistance to DRHGC wines can be somewhat overcome if the environmental benefits of DRHGCs are emphasised¹⁷.

So far, little work has been done regarding DRHGC wines produced in Italy. While the production of wines from DRHGCs remains uncommon in most of Italy, the region of South Tyrol (German, Südtirol; Italian, Alto Adige) has pioneered their use because many DRHGCs are suited to cold climates and mountain viticulture.

DRHGCs require fewer treatments with phytosanitary products, which is particularly critical in mountainous regions where mechanisation is difficult or impossible². In terms of climate, the region ranges from Zone I to III on the Winkler classification¹⁸. In addition, vineyards at higher altitudes (above 1100 m) are classified as Zone WI 1a and thus are considered unsuited for V. vinifera cultivars¹⁸ but may nonetheless suit cold-hardy DRHGCs.

The red V. vinifera cultivars that are most widely grown in South Tyrol are the international cultivars V. vinifera ‘Pinot nero’ (‘Blauburgunder’, ‘Spätburgunder’, ‘Pinot noir’), ‘Cabernet Sauvignon’, ‘Cabernet Franc’, and ‘Merlot’, and the local cultivars V. vinifera ‘Schiava’ (‘Vernatsch’, ‘Trollinger’) and ‘Lagrein’. A smaller number of producers have planted red DRHGCs, including Vitis sp. ‘Cabernet Cortis’ and ‘Regent’.

This study aims to determine the chemical and sensory profile of red DRHGC and V. vinifera wines from South Tyrol, Italy, to identify their most characteristic aspects. For that purpose, a strategy based on a combination of sensory analysis, using mRATA and projective mapping, and chemical analysis, using LC-MS/MS and HS-SPME-GC×GC-ToF/MS, was implemented.

Results and discussion

Anthocyanins

Red wines are defined by the presence of anthocyanins in free or bound form, which are the main pigments responsible for wine colour^19,20. The anthocyanin profile of DRHGC grapes and wine is, however, distinctively different to that of V. vinifera^2,21,22.

In particular, diglucoside anthocyanins are typically only found in DRHGCs. In addition, 3,5,7-O-triglucoside anthocyanins have been tentatively reported in V. amurensis²³. In this study, we tentatively characterised the structure of several different anthocyanins in DRHGC and V. vinifera red wines. These were identified based on retention time (RT), molecular ions ([M]⁺), main fragment ions, and UV–visible spectra absorbance relative maxima (λ-max) generated using a UHPLC-QqQ/MS instrument with a photodiode array (PDA) detector. For some of them, a sufficient separation and an acceptable peak shape could be achieved only applying the described chromatographic conditions, particularly 10% v/v formic acid (phase A, pH ~ 1.4). Although very high, this concentration of formic acid in the mobile phase was necessary to attain the results presented herein. Lower concentrations of formic acid (e.g. ≤5% v/v) did not allow good separations for some/most features which characterise the DRHGC wines (see below). Some species could not be detected at all because their peak was too “flattened” at higher pH, a typical phenomenon observed during anthocyanin analysis when an insufficient pH-modifier is added to the mobile phase. The use of a stationary phase able to withstand such low-pH conditions was therefore a requirement (see the experimental section). A total of 30 anthocyanins were detected in the wine samples studied, which included DRHGCs and V. vinifera wines, and 26 of these were tentatively identified. Based on their chemical structures (Fig. 1) and on current literature^24,25, the 26 anthocyanins that were tentatively identified were all derivatives of five anthocyanidins: delphinidin (Dp), cyanidin (Cn), petunidin (Pt), peonidin (Pn), and malvidin (Mv). Notably, regardless of the type of derivative (mono-glycosides, di-glycosides, tri-glycosides, acetylated, etc.), for the same type of substituted species and the type of LC stationary phase used (C18), the order of elution always followed exactly the presented aglycones order (Dp → Cn → Pt → Pn → Mv), which also helped in the identification of compound series with the same type of structural derivatisation/substitutions. The list of all identified compounds is summarised in Table 1.

Fig. 1: Tentative structures of anthocyanins detected in the DRHGCs and V. vinifera wines used in this study.

G glucose, diG diglucoside. Chemical structures are adapted from Wiczkowski et al.⁴², Xu et al.²⁵, Darnal et al.²⁴, Morata et al.²⁹, Wang et al.⁵⁰. Dp delphinidin, Cn cyanidin, Pt petunidin, Pn peonidin, Mv malvidin, G glucoside, diG diglucoside, R functional group.

Table 1 Tentative assignments of anthocyanins

All MS² spectra traces of all predicted anthocyanins are summarised in Supplementary Data S10. The generic profile of anthocyanins in two different samples is reported in Supplementary Fig. S2. The PDA traces were employed to identify pigment compounds according to their absorption at 520 nm (Table 1). The assignment of most compounds detected in MS¹ analysis was confirmed by target MS² analysis.

Diglucosides were tentatively identified. The MS spectrum of peak 1 showed the molecular ion [M]⁺ at m/z 627.1 and MS² fragments at m/z 465.1 ([M-162]⁺) and 303.1 ([M-162-162], [Dp]⁺). The maximum wavelength at 520 nm with an E₄₄₀/E_vis-max of 19% indicated that this compound was an anthocyanin 3,5-diglucoside and its aglycone was delphinidin (Dp) (m/z 303.1). The loss of two units of 162 amu indicated the presence of two glucosyl moieties linked to the aglycone. The presence of 3,5-diglucoside anthocyanins with similar substitution patterns in DRHGCs that are not present at quantifiable levels in V. vinifera is widely reported in the literature, thus supporting this hypothesis. Thus, this compound was tentatively identified as delphinidin 3,5-O-glucoside (Dp3,5G). Peak 3 (m/z 611.1 [M]⁺, 449.1 [M-162]⁺, 287.1 [M-162-162]⁺, [Cn]⁺), peak 6 (m/z 641.1 [M]⁺, 479.1 [M-162]⁺, 317.1 [M-162-162]⁺, [Pt]⁺), peak 8 (m/z 625.1 [M]⁺, 463.1 [M-162]⁺, 301.1 [M-162-162]⁺, [Pn]⁺), and peak 10 (m/z 655.1 [M]⁺, 493.1 [M-162]⁺, 331.1 [M-162-162]⁺, [Mv]⁺) showed the same fragmentation pattern as peak 1. Hence, they were tentatively identified as cyanidin-3,5-diglucoside (Cn3,5G), petunidin-3,5-O-diglucoside (Pt3,5G), peonidin 3,5-diglucoside (Pn3,5G), and malvidin 3,5-diglucoside (Mv3,5G), respectively. Peak 10 (m/z 655.1) was confirmed by the injection of the reference standard of Mv3,5G.

Monoglucosides were also tentatively identified. The MS spectrum of peak 7 presented a major signal at m/z 465.1 and the MS² spectrum showed one signal at m/z 303.1 that originated from the molecular ion by loss of 162 ([M-162], [Dp]⁺). The fragment at m/z 303.1 represented the aglycone of Dp. The maximum wavelength at 525 nm indicated that this compound was an anthocyanin. The E₄₄₀/E_vis-max value (%) of 41 indicated an anthocyanidin 3-O-glycoside type²⁶. Therefore, peak 7 was tentatively identified as delphinidin 3-O-glucoside (Dp3G). Peak 9 (449.1 [M]⁺, 287.1 [M-162]⁺, [Cn]⁺), peak 13 (479.1 [M]⁺, 317.1 [M-162]⁺, [Pt]⁺), peak 14 (463.1 [M]⁺, 301.1 [M-162]⁺, [Pn]⁺), and peak 17 (493.1 [M]⁺, 331.1 [M-162]⁺, [Mv]⁺) showed the same fragmentation scheme. Therefore, they were tentatively identified as cyanidin-3-glucoside (Cn3G), petunidin-3-O-glucoside (Pt3G), peonidin-3-glucoside (Pn3G), and malvidin-3-glucoside (Mv3G), respectively. Peaks 9 and 16 shared the same fragmentation pattern. The aglycone ion (m/z 287.1) indicated that both peaks corresponded to cyanidin derivatives as they had the same molecular ions [M]⁺ at m/z 449.1 and showed the same MS² spectrum fragmentation patterns, indicating that they could be isomers such as a glycoside containing galactosyl or pyranosyl units^27,28. Additionally, peak 9 and 16 showed maximum wavelength at 545 and 510 nm, respectively. Peak 16 was hypothesised to be a derivative of Cn3G, in which the difference of 162 amu from the molecular ion was assigned to a loss of glucose residue.

A number of pyroanthocyanins were also tentatively identified. Peak 20 possibly corresponded to vitisin B (VitB), which belongs to the pyranoanthocyanin pigment family and is derived from reactions between anthocyanins and acetaldehyde during fermentation and ageing^29,30,31. The mass spectrum of the pigment obtained by LC/DAD-MS in the positive ion mode showed a molecular ion [M]⁺ ion at m/z 517 and a fragment ion [M − 162]⁺ at m/z 355, corresponding to the loss of the glucose moiety.

Peak 21 was detected as vitisin A (VitA), which also belongs to the pyranoanthocyanin pigment family and is formed by chemical reactions between Mv3G and pyruvic acid during fermentation and ageing²⁹. The molecular ion [M]⁺ ion at m/z 561 and a fragment at m/z 399 tentatively confirmed the assignation as VitA²⁹.

These molecules have previously been reported by several authors^{32,33,34,35,36,37} and they have already been detected in DRHGCs and V. vinifera, in which differences between the accumulation of these anthocyanins were found.

We also tentatively identified several acetylglucoside conjugates of anthocyanins. The molecular ion of peak 19 was detected at m/z 507.1 and fragmented into m/z 303.1. The difference of 204 amu in the molecular ion could be attributed to the loss of glucose (162 amu) and an acetyl residue (42 amu); however, the monoglucoside was not observed. The E₄₄₀/E_vis-max value (%) of 25 indicated a glucosyl moiety and a 3,5-disubstituted anthocyanin²⁶. Hence, peak 19 was accordingly identified as delphinidin-3-(acetyl)-O-glucoside (Dp3acG). Following this hypothesis, peak 23 (491.1 [M]⁺, 287.1 [M-208]⁺, [Cn]⁺), peak 24 (m/z 521 [M]⁺, 317.1 [M-208]⁺, [Pt]⁺), peak 27 (m/z 505.1 [M]⁺, 301.1 [M-208]⁺, [Pn]⁺), and peak 30 (m/z 535.1 [M]⁺, 331.1 [M-208]⁺, [Mv]⁺) were tentatively identified as cyanidin 3-(acetyl)-O-glucoside (Cn3acG), petunidin-3-(acetyl)-O-glucoside (Pt3acG), peonidin-3-(acetyl)-O-glucoside(Pn3acG), and malvidin-3-(acetyl)-O-glucoside (Mv3acG), respectively. These fragmentation patterns were compared with the current literature^25,28,38. Acetylglucose conjugates of anthocyanins have already been reported in both DRHGC and V. vinifera red wines^39,40,41. Acylation of a monoglucoside occurs preferentially on the glucose moiety at the C6 position.

The presence of triglucoside anthocyanins in wine samples included in this study was investigated and their MS spectrum was compared with the MS spectrum of a red cabbage extract (Supplementary Data S10). The MS² analysis of peak 2 at 12.10 min revealed a molecular ion at m/z 773 [M]⁺ and fragment ions at m/z 611 [M-162]⁺, m/z 449 [M-162-162]⁺ and m/z 287 [M-162-162-162]⁺ in both mass spectra of a red cabbage extract and of a mixture of MonCabCort and MalFochCabCort wine samples. The neutral loss of 162 amu corresponded to the loss of a glucosyl moiety. Three neutral losses of 162 amu were detected within the structure, leading to the tentative identification of a cyanidin-diglucoside-glucoside (CndiGG), which is one of the major anthocyanins in red cabbage (Wiczkowski et al.⁴²). Therefore, considering the presence and co-elution of the peak 2 in both samples (red cabbage extract and a mixture of MonCabCort and MalFochCabCort wine), we could tentatively propose that these anthocyanin are sophoroside types (2-O-β-D-glucopyranosyl-α-D-glucopyranose). To summarise, based on MS² spectrum, retention time, and UV-Vis spectrum of peak 2 are consistent with the compound extracted from red cabbage. Based on the literature on red cabbage (Wiczkowski et al.⁴²), it was therefore assigned to CndiGG, presumably as 3O-sophoroside-5O-glucoside type.

Peak 4 (787.1 [M]⁺, 625.1 [M-162]⁺, 463.1 [M-162-162]⁺, 301.1 [M-162-162-162]⁺, [Pn]⁺), and peak 5 (m/z 817.1 [M]⁺, 655.1 [M-162]⁺, 493.1 [M-162-162]⁺, 331.1 [M-162-162-162]⁺ [Mv]⁺) also followed the cleaving processes found in the CndiGG. Hence, they were tentatively identified as peonidin diglucoside-glucoside (PndiGG), and malvidin diglucoside-glucoside (MvdiGG), respectively. These fragmentation patterns were compared with the current literature^23,42. The delphinidin-diglucoside-glucoside (DpdiGG) was only detected in red cabbage (m/z 789.1 [M]⁺, 627.1 [M-162]⁺, 465.1 [M-162-162]⁺, 303.1 [M-162-162-162]⁺ [Mv]⁺, Supplementary Data S10). Petunidin diglucoside-glucoside was not detected (at a limit of detection of 0.025 mg mL⁻¹, Supplementary Data S11).

Anthocyanins with more than two glycosidic residues have been already reported in the literature. Their sugar residues are usually linked as either 5,3,5′-triglucoside or 3,7,3′-triglucoside, and can contain glucose, galactose, rhamnose, and arabinose units⁴³. Zhu et al.²³ speculated that 3,5,7-O-triglucoside anthocyanins might be present in grapes from V. amurensis and its hybrids. In contrast, our results tentatively demonstrated that the molecular ion at m/z 773 (RT 12.10 min) was a CndiGG, thus agreeing with Wiczkowski et al.⁴² and Gonzali and Perata⁴⁴, in which the glycosyl moieties were bound to the 3-position on the C-ring or the 5- or 7-position on the A-ring. However, only trace amounts were found in the DRHGC wine samples included in this study. In summary, diglucoside-glucoside of Pn and Mv were found in a mixture of MonCabCort and MalFochCabCort wine samples, while Pt and Dp diglucoside-glucoside were not detected and were therefore considered to likely be present only in trace amounts. The elution order of Cn (peak 2, m/z 773, RT = 12.10), Pn (peak 4, m/z 787, RT = 15.2), and Mv (peak 5, m/z 817, RT = 15.9) tentatively further confirmed our assumption since elution order is determined by the number of hydroxyl groups, their degree of methoxylation (De < Cy < Pe < Pn < Mv), the number of glucosides, and their acylation pattern (GLU2 < GLU < GLU–AC < GLU2–CO < GLU–CA < GLU–CO)⁴⁵. Indeed, diglucoside-glucoside anthocyanins were eluted earlier than all the other anthocyanins detected in the study samples²³. Additionally, comparable fragmentation patterns between the wine mixture and the red cabbage extract were found. Hypothetical chemical structures of diglucoside-glucoside anthocyanins and the hypothetical fragmentation pathway of malvidin diglucoside-glucoside (m/z 817.1) are reported in Supplementary Data S10.

Finally, p-coumaroylated diglucoside anthocyanins were tentatively identified. Distinctive anthocyanin features between red wines from DRHGC and V. vinifera wines were identified at high retention times (see Table 1). Evaluation of the chromatographic and spectral data (UV-Vis/PDA, and MS/MS) hinted at two tentative assignments, with identical exact masses but different molecular structures: (caffeoyl)rutinoside or (p-coumaroyl)3,5-diglucoside derivatives (of all 5 basic anthocyanidin structures). The presence of anthocyanin rutinosides has been reported in other plants or food matrices, including Iris L²⁵., red-skinned and purple-fleshed potatoes (Solanum tuberosum L.)⁴⁶, ripe fruits of Garcinia indica (Thouars) Choisy²⁸, Geraniaceae Juss. and Lamiaceae Martinov families⁴⁷, and the flowers of Petunia reitzii L.B.Sm. & Downs⁴⁸. We could not locate any reports that provided such detailed anthocyanin rutinosides profiling for V. vinifera or DRHGC wines, but one previous study has identified their presence in wine produced from V. davidii (Rom.Caill.) Foex. cultivars⁴⁹. A representative MS² spectrum of an anthocyanin fragmentation pattern is shown in Supplementary Fig. S3. However, given tentative assignment given to the analogous species in other wines⁵⁰, this compound with m/z 801.1 eluting at RT 49.4 min should be assigned to malvidin-(p-coumaroyl)-3,5-O-diglucoside (Supplementary Fig. S3).

The tentative identification of these compounds followed a process of elimination of other potential candidate compounds and represents the only hypothesis able to explain all observations (LC-MS, MS², and UV-Vis features). For example, a strong UV-Vis absorption band between 300 and 400 nm was observed, which was comparable in intensity to the Vis band at >500 nm, a characteristic which is usually attributed to glycosylated anthocyanins containing an esterified hydroxycinnamic acid. In addition, the MS² fragmentation pattern did not show a transition compatible with a malvidin-linked single esterified (caffeoyl)rutinoside unit, or a (p-coumaroyl)diglucoside unit (isomeric structures). However, the cleavage of the ester link does not occur in relation to observed MS/MS fragmentation peaks⁵⁰. Therefore, (p-coumaroyl-3,5-diglucoside) moiety was preferred, also considering the above cited literature reports⁵⁰.

The MS spectrum of peak 22 presented a parent ion at m/z 773.1 and the MS² spectrum showed three fragment ions (at m/z 611.1, 465.1, 303.1), corresponding to the losses of 162 ([M-162]⁺), 146 ([M-162-146]⁺), 162 ([M-162-146-162]⁺) amu, respectively. The UV-Vis spectrum of this compound showed two peaks of absorbance maximum at 312 and 531 nm with an E₄₄₀/E_vis-max of 34%. The maximum wavelength at 531 is characteristic of anthocyanins, and an additional absorbance maximum at 312 nm is characteristic of a caffeic acid acylation. Given that the weight of the lost glycosyl residue is 162²⁵. The remaining MS² fragment ion was observed at m/z 611. The second fragment ion at m/z 465.1 resulted from the loss of the entire (p-coumaroyl)glucosyl moiety ([M—162-146]⁺. Indeed, the loss of 308 amu (162 + 146 amu) indicates the simultaneous presence of both glucose and a p-coumaroyl unit linked at the same position of the aglycone, thus contributing to the evidence that this compound might have been a (caffeoyl)-rutinoside derivative. The third ion generated in the MS² analysis (m/z 303.2) corresponded to the anthocyanidin delphinidin and resulted from the loss of all the substituents (470 amu) from the molecular ion. Peak 22 was accordingly identified as delphinidin-(p-coumaroyl)-3,5-O-diglucoside (Dp3,5CD). Peak 25 (757.1 [M]⁺, 595.1 [M-162]⁺, 287.1 [M-308-162]⁺, [Cn]⁺), peak 26 (m/z 787.2 [M]⁺, 625.1 [M-162]⁺, 479.1 [M-162-146]⁺, 317.1 [M-162-146-162]⁺ [Pt]⁺), peak 28 (m/z 771.2 [M]⁺, 609.1 [M-162]⁺, 301 [M-308-162]⁺, [Pn]⁺), and peak 29 (m/z 801.1 [M]⁺, 639.1 [M-162]⁺, 493.1 [M-162-146]⁺, 331.1 [M-162-146-162]⁺ [Mv]+) also followed these cleaving processes. Hence, they were tentatively identified as cyanidin-p(coumaroyl)-3,5-diglucoside (Dp3,5CD), petunidin-p(coumaroyl)-3,5-diglucoside (Pt3,5CD), peonidin-p(coumaroyl)-3,5-diglucoside (Pn3,5CD), and malvidin-p(coumaroyl)-3,5-diglucoside (Mv3,5CD), respectively. These fragmentation patterns were compared with the current literature^25,28,50. The elution order of Dp3CD (peak 22, m/z 773, RT = 39.9, (Supplementary Fig. S4a, red)), Cn3,5CD (peak 25, m/z 757, RT = 43.7, (Supplementary Fig. S4b, black)), Pt3,5CD (peak 26, m/z 787, RT = 45.5, (Supplementary Fig. S4d, blue)), Pn3,5CD (peak 28, m/z 771, RT = 48.7, (Supplementary Fig. S4e, purple)), and Mv3,5CD (peak 29, m/z 801, RT = 49.4, (Supplementary Fig. S4c, green)) tentatively further confirmed our assumption^45,50 as reported in Supplementary Fig. S4.

A rutinoside unit linked to C3 of the C-ring and a glucose bound to C5 of the A-ring has been already reported²⁵. Therefore, it was initially hypothesised that these anthocyanins, namely peak 22, 25, 26, 28, and 29, were bound to a rutinoside unit attached to C3 of the C-ring and that one sugar residue, possibly the rhamnose and not the glucose, was acylated by a caffeic acid. For a few anthocyanin types, these acids are found to attach to the 6-hydroxyl and 4-hydroxyl group of the second sugar⁴³. The fact that previous studies of natural products have demonstrated the presence of anthocyanins with similar substitution patterns, and specifically with rutinoside, supports the identification of rutinoside anthocyanins in the present study. However, the interpretation of these peak with the esters with p-coumaric acid of the diglycosidic anthocyanins was eventually preferred, as evidence showed that the cleavage of the ester bond is not associated with peaks in MS/MS analysis, therefore the 162 amu loss was most likely due the loss of a glycoside unit instead.

Peak 22 (m/z 773.1, RT = 39.9), which was tentatively identified as Dp3,5CD, shared the same parent ion as CndiGG at m/z 773.1 (as described earlier in the subsection ‘Anthocyanins’ in Results and discussion). The different retention times (39.9 min and 12.1 min respectively), the different fragmentation patterns, and the presence of a visible maximum wavelength at 531 and absorbance maximum at 312 tentatively suggested that the peak 22 could be Dp3,5CD. Also, peak 12 showed a signal at m/z 773.1 (RT = 20.4 min), which was present only in V. vinifera, but this could not be identified from the current literature.

Additionally, UV-Vis data were used to characterise and to provide a tentative confirmation of structure, based on the substitution pattern on the anthocyanin backbone. The typical UV-Vis spectrum of an anthocyanin is characterised by two peaks, the first of which is observed at wavelength of around 260–280 nm (UV region), and the second is observed in the visible region at around 510–520 nm. Additionally, a hump in the range of 400–450 nm is also observed, the size of which is dependent on the number of sugar moieties attached to the anthocyanin structure. A peak in the range of 310–340 nm indicates if the sugar moiety is acylated²⁶.

Previous studies have shown that the ratio (expressed as a percentage) between E₄₄₀/E_vis-max can be used for preliminary identification of anthocyanin types^25,26,51. An E₄₄₀/E_vis absorptivity ratio of 29–35 (%) indicates a glucosyl moiety, whereas a lower value of 15–24 (%) is observed for a diglucosyl moiety. A glycosidic substitution at position 3 of the anthocyanin backbone produces a ratio two times higher than when the substitution is found at position 5 or both 3 and 5. A pelargonidin substituted at 3-position has a ratio between 0.38 and 0.46, while in the case of 5-substituted and 3,5-disubstituted, the ratio is between 0.18–0.22²⁶. In contrast, a 4-substituted anthocyanin will show a higher absorbance at 440 nm²⁶.

Therefore, based on a review of the relevant literature²⁶, Table 1 provides tentative assignments of the compounds found in this study.

The anthocyanin profiles of DRHGC and V. vinifera wines were analysed using Multiple Factor Analysis (MFA) and are shown in Fig. 3g. Individuals are shown in Fig. 3b. The MFA showed that each wine sample was characterised by a unique anthocyanin profile and that a clear separation between DRHGC and V. vinifera wines was achieved. Anthocyanins contributed 9.32% to the first dimension, 19.64% to the second, and 7.81% to the third.

Tentative identifications of all observed anthocyanins are reported in Table 1. Differences in the anthocyanin content are reported between DRHGC and V. vinifera wines, but also between DRHGC wines.

Mv3G (m/z 493.1), VitA (m/z 561.1), VitB (m/z 517.1), Mv3acG (m/z 535.1), Pt3acG (m/z 521.1), Dp3acG (m/z 507.1), and Pn3G (m/z 463.1, RT = 27.24) were significantly higher in CabSauv sample from Winery 8 (1-way ANOVA, p < 0.05) than the other samples.

The highest anthocyanin presence was observed in the MonCabCort wine sample from Winery7, which included a very high abundance of both monoglucosides and diglucoside anthocyanins, namely Dp3G (m/z 465.1), Cn3G (m/z 449.1-1, RT = 20.29), Pt3G (m/z 479.1), and Mv3,5 G (m/z 655.1), Dp3,5 G (m/z 627.1), Cn3,5 G (m/z 611.1), Pt3,5 G (m/z 641.1), and Pn3,5 G (m/z 625.1). Furthermore, Pn3,5CD (m/z 771.1) was higher in the MonCabCort sample from Winery7 than in any of the other wine samples. According to the information supplied by the winemaker, this sample was a blend between ‘Monarch’ and ‘Cabernet Cortis’ varieties.

Dp3,5CD (773.1-1, RT = 39.09 min) and Pt3,5CD (m/z 787.1) were higher in MalFochCabCort samples from Winery7 than the other samples. According to the information supplied by the winemaker, this sample was blend between Vitis sp. ‘Maréchal Foch’ and ‘Cabernet Cortis’.

CabCort from Winery 3 showed the highest content of Mv5CR (m/z 801.1). According to the information obtained by the winemaker, this sample was a monovarietal ‘Cabernet Cortis’ wine.

The relative abundance of 22 anthocyanins out of 30 detected in the wine samples are summarised into five groups: acylated (ac) (Dp3acG, Cn3acG, Pt3acG, Pn3acG, Mv3acG); p-coumaroylated diglucosides (CD) (Pt3,5CD, Dp3,5CD, Pn3,5CD, Mv3,5CD); diglycosylated (DiG) (Dp3,5G, Cn3,5G, Pt3,5G, Pn3,5G, Mv3,5G), glycosylated (MoG) (Dp3G, Cn3G, Pt3G, Pn3G, Mv3G), and pyranoanthocyanin (pyroA) (VitA, VitB) (Fig. 2).

Fig. 2: Total content and proportion of acylated (ac), p-coumaroylated diglucosides (CD), diglycosylated (DiG), glycosylated (MoG), and pyranoanthocyanin (pyroA) anthocyanins in each wine sample.

The total content is expressed as a sum of each individual anthocyanin detected in this study as reported by Xu et al.²⁵. Ac expressed the sum of Dp3acG, Cn3acG, Pt3acG, Pn3acG, Mv3acG; CD expresses the sum of Pt3,5CD, Dp3,5CD, Pn3,5CD, Mv3,5CD; DiG expresses the sum of Dp3,5G, Cn3,5G, Pt3,5G, Pn3,5G, Mv3,5G; MoG expresses the sum of Dp3G, Cn3G, Pt3G, Pn3G, Mv3G, pyroA expresses the sum of VitA and VitB. The anthocyanins are organised according to their tentative assignment (Table 1). The unknown compounds and the trace amounts of triglucosides are not included in this graph.

The MonCabCort samples showed the highest abundance of anthocyanins, followed by the MalFochCabCort, CabCort and CabSauv wine samples. It is worth noting that ‘Cabernet Sauvignon’ wine, as a V. vinifera cultivar, was the most characterised by the total abundance of acylated (acG), glycosyl anthocyanins (MoG), and pyranoanthocyanin (pyroA), whereas p-coumaroyl diglucoside anthocyanins (CD) and diglycosyl anthocyanins (DiG) mainly characterised DRHGC wines (Fig. 2). No significant differences in the total abundance of CR between DRHGC wines were found. The results are supported by 1-way ANOVA (p < 0.05, Supplementary Data S9). In summary, the ANOVA results showed that the MoG and DiG anthocyanins were most abundant in the blend between ‘Monarch’ and ‘Cabernet Cortis’ cultivars (Winery7_MonCabCort).

The differences in DiG content between the blend (Winery7_MonCabCort) and the monovarietal (Winery3_CabCort) samples could be a result of the different anthocyanin profiles of different grape cultivars. The profile of each cultivar is specific to the genotype of that cultivar^12,52,53. For example, delphinidin derivates are the most abundant anthocyanin in ‘Regent’³⁸. It is possible that the higher content of DiG anthocyanins observed in Winery7_MonCabCort than in the monovarietal sample wine (Winery3_CabCort) could be partially due to the presence of ‘Monarch’ rather than ‘Cabernet Cortis’ in the wine. Indeed, a comparative study between six grape cultivars showed that the total diglycosylated anthocyanins were higher in ‘Monarch’ than ‘Cabernet Cortis’⁵³. Furthermore, the distribution of individual anthocyanins depends on different factors, such as growing conditions, winemaking techniques employed, and bottle ageing⁵⁴.

Previous studies^11,53 reported that the most abundant anthocyanins in ‘Cabernet Cortis’ grapes were Mv3,5G, followed by Pt3,5G, Dp3G, Pn3,5G, Mv3G, and Dp3,5G. Regarding the DiG, our findings partially agreed with these results, since the most abundant anthocyanin was Mv3,5G, followed by Pt3,5G, Dp3,5G, Pn3,5G, and Cn3,5G in the MonCabCort sample. A similar anthocyanin distribution pattern was observed in the monovarietal ‘Cabernet Cortis’ sample (Winery3_CabCort). Theoretically, Mv, which is bound to a glycosyl moiety, is the most stable, followed by Pn, Pt, Cn, and Dp, since the stability to oxidation of anthocyanins decreases as the number of hydroxyl substituents on the B-ring increases¹¹. Hydroxylation contributes to colour stability, hue, and increased blueness, whereas methylation contributes to the increased redness⁵⁴. In addition, diglucosides generally have a higher stability to singlet oxygen than monoglucosides¹¹. An interrelationship between colour and anthocyanin content has been reported by different authors^46,55. Mv is associated with reddish purple colour⁵⁵, and the highest positive a* value (green-red colours) was observed in the Winery7_MonCabCort and the lowest in Winery3_CabCort, suggesting that the most reddish sample was MonCabCort (1-way ANOVA, p < 0.05; Supplementary Data S9)^56,57,58. In general, Pt, Pn, and Mv showed greater stability to oxidation than Cn and Dp¹¹. Additionally, when the hydroxylation occurs on the ortho-position, such as Dp, the susceptibility to oxidation increases¹¹.

Hence, as the abundance levels of Mv and Pt were higher than Dp in the MonCabCort sample compared to the other wine sample, it is suggested that the MonCabCort sample might be more stable to oxidation⁵⁴.

The presence of anthocyanin rutinosides was hypothesised, although the more likely explanation was considered to be that these compounds are anthocyanin p-coumaroyl diglucosides. Studies on anthocyanins containing rutinosides in grape wine matrices are lacking in the scientific literature, though their presence in grapes⁵⁹, V. davidii wines⁴⁹, and in fruit wines^60,61 has previously been discussed. Additionally, our findings provide new information on the mass spectral and UV-Vis data of Mv3,5CD, Pt3,5CD, Dp3,5CD, Pn3,5CD, and Cn3,5CD anthocyanins in a wine environment. It is widely recognised that the properties of anthocyanins, such as colour expression, are significantly influenced by both their chemical structure and by wine pH⁶². Typically, anthocyanins exhibit greater stability in acidic environments at lower pH values compared to alkaline solutions with higher pH values⁵⁴. In a wine environment with a pH of 3.5, anthocyanins exit in the flavylium form. However, to date, few studies have been reported on rutinoside anthocyanins in wine matrices and, therefore, on their behaviour during winemaking with either DRHGC or V. vinifera grapes. Our hypotheses are based on published works on other plant matrices.

Since an evaluation of the potential in vitro antimicrobial activity of the anthocyanin rutinoside compounds detected in DRHGC wines was not conducted in this study, a comparison with the current literature was made. Wegener and Jansen⁶³ reported that anthocyanins play a role in the resistance of coloured potato cultivars to soft rot caused by Pectobacterium carotovorum (Pc). Purple-fleshed potato cultivars in particular produce more secondary metabolites that are involved in plant defence mechanisms than white or yellow-fleshed cultivars. The expression of anthocyanin 5-O-glucosyl-transferase improved the resistance of transgenic potato lines to Pc. Among the three enzymes tested by Wegener and Jansen⁶³ (phenylalanine ammonia-lyase, polyphenol oxidase (PPO), and peroxidase (POD)), only PPO activity was higher in the coloured potato cultivars. This finding highlights the significant role of PPO in resistance expression and underscores the importance of the initial phase of plant defence, specifically the release of phenols and their oxidation by phenolases.

The higher content of anthocyanin rutinosides in DRHGC wines might suggest that these compounds could be involved in vine disease defence, a potential challenge for further research. Indeed, Kröner et al.⁶⁴ reported the involvement of the phenylpropanoid pathway in stable, multi-stress quantitative resistance. This is likely due to its ability to simultaneously generate numerous active secondary metabolites, each contributing to direct and/or indirect defence mechanisms.

Proanthocyanidins

Condensed tannins (PAC) are a class of polyphenolic compounds present in the solid parts of the grapes^65,66. Macrocyclic (cyclic oligomeric or crown) PAC (c-PAC) are of increasing interest in the literature because their physiochemical proprieties are still being investigated. The concentration of PAC and c-PAC varies across grape varieties⁶⁷ and can be influenced by winemaking conditions^65,68,69.

The list of observed proanthocyanidins is reported in Supplementary Data S4 and S5 for all wines investigated in this study. Additionally, percentage ratios (%) of cyclic tetrameric procyanidin (%C-4), cyclic tetrameric prodelphinidin with one (epi)gallocatechin, three (epi)catechins (%C-4-OH), and cyclic pentameric procyanidin (%C-5) were calculated as concentration-normalised parameters (equation reported by Darnal et al.⁶⁵). This has been proposed in previous literature as a wine authenticity indicator to investigate different wine origins^52,70. For the studied wines, it was hypothesised that different levels of PAC characterised DRHGC and V. vinifera wines.

The PAC profile of DRHGC and V. vinifera wines was investigated as part of the MFA (Fig. 3d). Individuals are shown in Fig. 3b. PAC contributed 8.80% to the first dimension, 2.81% to the second, and 12.49% to the third (Fig. 3a). A separate principal component analysis (PCA) extracted from the MFA is shown in Supplementary Fig. S6c, d. This suggests that these wines are characterised by a unique PAC profile. However, the DRHGC and V. vinifera wines were not well separated, suggesting that the differences between cultivars is more important than that between DRHGC and V. vinifera. The percentage ratios, %C-4, %C-4-OH, and %C-5 are projected as supplementary variables in MFA, because they are derived parameters⁶⁵.

Fig. 3: Overall MFA of red wine chemical and sensory data.

a Partial axes, b individuals coloured by grape type, c quantitative variables for basic oenological parameters, d quantitative variables for PAC, e quantitative variables for volatiles (cos2 = 0.5), f quantitative variables for polyphenols (cos2 = 0.5), g quantitative variables for anthocyanins and CIELab, h quantitative variables for sensory analysis. a. aroma, f. flavour.

Intriguingly, c-PACs as c-tetramer (m/z 1153) and c-pentamer-1 (m/z 1441-1) showed a high contribution to the separation of DRHGC samples from Winery 3 and Winery 7 along the positive direction of Dim.1. These DRHGC wines have unique c-PAC profiles, and this suggests that there are characteristic differences between DRHGC wines. Hence, c-tetramer and c-pentamer-1 were higher in CabCort samples from Winery 3 and these observed c-PACs were lower in MonCabCort and MalFochCabCort samples from Winery 7. No statistical differences between CabCort samples from Winery 3 and CabSauv samples from Winery8 were observed. In contrast, c-OH-tetramer (m/z 1169) was higher in MonCabCort and MalFochCabCort samples from Winery 3, since c-OH-tetramer showed negative values on Dim.1 and was opposite to c-tetramer and c-pentamer-1.

The dimers (m/z 579), OH-dimers (m/z 595), trimers (m/z 867), and tetramers (m/z 1155) were not significantly different between DRHGC and V. Vinifera samples. In contrast, OH-trimers (m/z 867) showed the highest content in DRHGC samples from Winery 3 (MonCabCort and MalFochCabCort). In general, pentamer (m/z 1443) and OH-tetramer (m/z 1171) were higher in CabCort from Winery3 and CabSauv from Winey 8 than the other samples. These differences were reported in Supplementary Data S9.

Regarding the percent ratios, %C-5 showed positive values on Dim.1, hence, they had the highest values in CabCort from Winery3 and CabSauv from Winery8 and these ratios had the same behaviour as the former ones. In contrast, %C-4-OH, which represented the ratio for the only identified cyclic prodelphinidin, had negative value on Dim.1 and they were higher in MonCabCort and MalFochCabCort samples from Winery 7. Interestingly, the %C-4-OH was anti-correlated to %C-5 (Fig. 3d), in a similar way to what was reported by Darnal et al.⁶⁵. The lowest content of %C-4 was detected in the Winery7_MonCabCort sample, followed by Winery3_CabCort, Winey8_CabSauv, and Winery7_MalFochCabCort. The results are supported by 1-way ANOVA (p < 0.05, Supplementary Data S9).

These ratios depend on the abundance of both cyclic compounds and their non-cyclic congeners. Indeed, the ratios are inversely proportional to the abundance of the latter, which is the reason for the higher values of %C-4-OH observed in Winery7_MonCabCort and Winery7_MalFochCabCort wines, and of %C-5 in Winery3_CabCort and Winey8_CabSauv. It is worth noting that %C-4-OH was higher in the blend samples, while %C-5 was higher in the monovarietal wines. The differences in trends observed for proanthocyanidins in wines are influenced by vinification procedures. In summary, the type of fermentation (alcoholic fermentation with or without malolactic fermentation)⁶⁵, maceration duration, and fermentation temperature⁷¹ affected the proanthocyanidin extraction and consequently red wine phenolic composition in another study. Several researchers have also found that the degree of polymerisation of proanthocyanidins varies significantly based on both the grape variety and the specific fraction of the grape being examined^40,71. It is worth noting that the mean degree of polymerisation undergoes significant changes during grape maturation as well⁷¹. Furthermore, during wine storage and ageing, modifications in the content of proanthocyanidins are reported. These alterations primarily result from a decrease in proanthocyanidin concentration due to their precipitation and a reduction in the mean degree of polymerisation, driven by subunit cleavage⁷¹.

Other phenolic compounds

The full MS and fragmentations of the phenolic compounds from our analysis were compared with the current literature reports and were tentatively identified.

Ions x.140 (m/z 397, RT = 10.7 min), x.72 (m/z 173, RT = 5.9 min), x.18 (m/z 191, RT = 1.4 min), x.40 (m/z 331, RT = 2.8 min), and x.83 (m/z 111, RT = 6.9 min) were statistically higher in the V. vinifera wine (Winery8_CabSauv) than the DRHGC samples. Ion x.140 was observed to have fragments ions at m/z 317, 289, 231, 165 and a molecular ion at m/z 397 ([M-H]⁻) in the MS² data. Based on Sobeh et al.⁷², this compound could tentatively be identified as (epi)catechin-5-O-ethyl 7-O-sulfate. However, we were not able to confirm the positions of the ethyl and sulfate groups substituent. Ion x.72 had a parent ion at m/z 173 and an MS² fragment ion at m/z 111, hence, it was tentatively identified as a shikimic acid derivative. Additionally, x.18 showed a parent ion at m/z 191 and a MS² fragment ion at m/z 111, hence, it was tentatively identified as a quinic acid derivative⁷³. Ion x.40 was characterised by m/z 331 and yielded a MS² fragmentation pattern composed of ions at m/z 169, 151, and 125. It was hypothesised to be a galloyl glucoside isomer, since the presence of a peak at m/z 331 indicated losses of a galloyl group (151 Da), m/z 169 loss of a hexosyl moiety (162 Da) and 125 loss of CO2 (44 Da), which indicated gallic acid fragments⁷⁴. Ion x.83 showed a peak at m/z 111 that was found to be strongly co-eluting with m/z 623, 311, 149, 179, which was hypothesised to be a fragment of trans-caftaric acid⁷⁵.

Ions x.129 (m/z 179 RT = 9.6 min), x.3 (m/z 205 RT = 1.0 min), and x.150 (m/z 119 RT = 11.7 min) were statistically higher in the CabCort sample from Winery3 than in the other samples. Ion x.129 was identified as caffeic acid and had a molecular ion at m/z 179 ([M-H]⁻). It was observed to have the MS² fragments m/z 161 (loss of water [M-H-162-18]⁻) and m/z 135 (carbon dioxide [M-H-162-44]⁻)⁷⁶. Ion x.3 showed an ion at m/z 205, a MS² fragment ion at m/z 125 and further peaks at m/z 289 and 145 were found in full MS mode. A possible identification was determined by comparing our findings with the current literature. Sun and Miller⁷⁷ reported that an ion at m/z 125 was observed in the MS/MS product ion spectra of modified catechins, such as gallocatechin and epigallocatechin. Furthermore, two abundant fragment ions at m/z 205 and 145 were reported from the fragmentation of catechin and epicatechin. Hence, ion x.3 was tentatively identified as a derivative of a catechin or a modified catechin^77,78. Ion x.150 showed ions at m/z 119 and m/z 163 in full MS mode. This 163 → 119 fragmentation pattern is typical of coumaric acid⁷⁹. Without further information regarding this compound, it was characterised as a coumaric acid derivative.

The DRHGC wines (Winery7_MonCabCort) were characterised by x.16 (m/z 305, RT = 1.3 min), and x.168 (m/z 181, RT = 15.3 min). x.16 produced a [M–H]⁻ ion at m/z 305, a MS² fragment ion at m/z 125, and further peaks at m/z 275, 289, 221, 203⁸⁰. This compound was identified tentatively as a derivative of a catechin. A peak at m/z 289 was also found to co-eluting at 1.3 min, thus confirming our hypothesis. The x.3 ion, which was found in Winery3_CabCort, showed a similar fragmentation pattern, therefore x.3 (RT = 1 min) and x.16 (RT = 1.3 min) were tentatively identified as a derivative of a catechin^77,78. x.168 showed a [M − H]⁻ at m/z 181 and MS² fragment ions at m/z 153 and 109. The last two were reported to be the typical fragmentation pattern of protocatechuic acid⁸¹. Consequently, it was speculated that x.168 was a derivative of protocatechuic acid.

The polyphenol profiles of DRHGC and V. vinifera wines were analysed using MFA and are shown in Fig. 3f (filtered using cos2 = 0.5 for readability; an unfiltered version is found in Supplementary Fig. S5b). Individuals are shown in Fig. 3b. The MFA showed that each wine sample was characterised by a unique polyphenol profile and that a clear separation between DRHGC and V. vinifera wines was achieved. The polyphenols (LC-MS) dataset contributed 12.8% to the first dimension, 17.9% to the second, and 19.8% to the third.

Our findings are broadly in agreement with that of the literature, which shows that the polyphenol profile of DRHGCs is distinctive, but differs between cultivars².

Volatile compounds

The list of observed volatile compounds is reported in Supplementary Data S8 for all wines investigated in this study. In total, 142 volatile compounds were identified in the wine samples. The volatile compounds shown in Fig. 3e (MFA, filtered using cos2 = 0.5 for readability; an unfiltered version is found in Supplementary Fig. S5a) and Supplementary Fig. S6g, h (PCA) were selected based on cos2 determined by the MFA.

The volatile compounds that were most characteristic of the V. vinifera wine (Winery8_CabSauv) were oxo-acetic acid (LXVI), hexanoic acid (CXIII), linalyl acetate (CXXIII), and 2-hydroxypropanoic acid ethyl ester (ethyl lactate; CXLII). Acetic acid is described as being pungent, sour, or vinegary⁸², hexanoic acid is associated with the sweat aroma descriptor⁸³, linalyl acetate is described as floral, sweet, citric, minty and slightly caraway-like⁸⁴, and ethyl lactate as fruity, buttery, and butterscotch⁸².

The DRHGC wines (Winery3_CabCort) were more characterised by diethyl butanedioate (ethyl succinate; LXXXII), ethyl 3-methylbutyl succinate (ethyl isopentyl succinate; LXXXIII), benzeneacetaldehyde (LXXVIII), (2E)-3-(3,4-dimethoxyphenyl)-2-propenoic acid ((E)-3,4-Dimethoxycinnamic acid; XXXV), methylenecyclobutane (XCVII), ethyl 9-decenoate (CV), phenylethyl alcohol (CXXXIX). Notably, ethyl succinate (as well as other esters) is reported to give fruity aromas⁸⁵, benzeneacetaldehyde has fatty and floral odour descriptors⁸⁶, ethyl 9-decenoate contributed to the flower-like aroma⁸⁷, phenylethyl alcohol contributed to the floral, spicy, herbal, and rose notes of the aroma profile^85,86.

Ethyl 2-hexenoate (XXV), 3,6-dihydro-4-methyl-2-(2-methyl-1-propenyl)-2H-pyran (nerol oxide; XXVI), trans-1-methyl-4-isopropenyl-2-cyclohexen-1-ol (XVI), benzaldehyde (LXIX), and α-terpineol (CXV) were higher in DRHGC wines (Winery7_MonCabCort). Notably, benzaldehyde contributed to the almond note and α-terpineol is reported to give a floral aroma⁸⁸.

As expected, our results found compounds typical of grapes related to ‘Cabernet Sauvignon’. These include 3-Hexen-1-ol (XLVII and XLVIII), which has grass and vegetal aromas, and eucalyptol (CVIII), which has mint and Eucalyptus aromas. Since all wines contained ‘Cabernet Sauvignon’ related grapes, these did not differ significantly between DRHGC and V. vinifera wines.

The ANOVA results were reported in the Supplementary Data S9, and they are used to confirm the above assumption.

The volatile profiles of DRHGC and V. vinifera wines were analysed using MFA and are shown in Fig. 3e. Individuals are shown in Fig. 3b. The MFA showed that a unique volatile profile characterised each wine sample and that a clear separation between DRHGC and V. vinifera wines is achieved. The volatile (GC×GC-MS) dataset contributed 13.6% to the first dimension, 14.8% to the second, and 18.5% to the third.

As with the polyphenol profile (section ‘Other phenolic compounds’ in Results and discussion), it seems likely that the volatile profile varies between DRHGCs and V. vinifera, but that the precise differences in volatilome will also vary between cultivars².

Sensory

The MFA was used to graphically visualise the samples and sensory attribute distributions. Figure 3h shows that the first two factors explained 53% of the variation in sensory data. mRATA sensory analysis contributed 13.01% to the first dimension, 5.35% to the second, and 9.27% to the third, while projective mapping sensory analysis contributed 12.3% to the first dimension, 7.41% to the second, and 15.6% to the third. Again, individuals are shown in Fig. 3b. The definition of each sensory descriptor is given in Supplementary Data S1. Results are collected in Supplementary Data S3 and S9.

Figure 3h shows the correlation between sensory variables and between sensory variables and dimensions. These expressed the perception of each attribute given by the panel (12 people, aged from 25–35 years old, see subsection ‘Sensory analysis’ in Methods) using a combination of mRATA and projective mapping methods.

mRATA was based on an intensity scale of each attribute perceived by the panel, and projective mapping showed how the panel located wine samples on mapping sheets (size 52 × 35 cm). The green bell pepper aroma and flavour attributes (known as herbaceous or vegetative attributes⁸⁹) mainly showed negative values on both Dim.1 and Dim.2. Dim.1. opposed the vegetative notes to woody attributes, such us oak-aroma and woody-flavour; indeed, woody attributes showed positive value on Dim.1. Also, tobacco-aroma (herbaceous or vegetative - dried) and liquorice-flavour (spicy) were located in the same position, thus showing a positive correlation with the woody group. Mint-aroma (herbaceous or vegetative – fresh), liquorice-aroma (spicy), and pungent-aroma (alcohol) had higher values on Dim.2 than all the other sensory attributes studied, and they were opposite to the perceived fruity notes such as cherry-aroma (tree fruit), red fruit-flavour (berry), strawberry-aroma (berry), and jam-aroma (dried). Similarly, the mouthfeel attributes such as sourness, warmness, astringency, and bitterness were associated with positive and negative values on Dim.1 and Dim.2.

Overall quality, which was expressed by the panel as an intensity scale and combined all three aspects of the sensory analysis (visual, olfactory, and gustatory), is considered an objective evaluation of the overall quality of the product⁷⁰. The panel preferred wines that were more characterised by woody attributes, such us woody-flavour and spicy notes e.g. black pepper-flavour. The flavour attributes that were most associated with the overall quality were bitterness and saltiness. The wine characterised by these sensory features was CabCort from Winery 7.

The position for each wine in the projective mapping method, which was expressed as the X and Y coordinates of each sample on the mapping sheet marked by each panellist, is shown on the MFA (Fig. 3h). The most plausible interpretation of the projective mapping method using MFA⁹⁰ was that the panellist relied on different characteristics of samples to assess their similarities and differences. In other words, they used divergent criteria or put different emphasis on different aspects of the samples. For example, panellist “632” primarily used green bell pepper-aroma and flavour (vegetative attributes) to place wines on the X axis of the mapping sheet and describe the wine samples, and the Y axis to organise the wine samples according to the woody and spicy attributes such as liquorice-aroma, liquorice-flavour, and oak-aroma. So, in the case of panellist “632”, they mainly organised and grouped the sample according to vegetative, woody, and spicy attributes. The same hypotheses may be constructed for all the other panellists.

Overall, the V. vinifera wine was ‘warmer’ and more astringent than the other wine samples, and the fruity notes (strawberry-aroma and jam-aroma) were more prominent in this wine sample (Winery 8 CabSauv). These results matched the information on the label, even if the fruity notes were not statistically different from the other samples. According to the product label, this wine was fermented with skin contact and temperature-controlled conditions in stainless steel tanks. After malolactic fermentation, it was aged for 12 months in barrels, followed by at least another 12 months of bottle storage.

Interesting similarities and dissimilarities between DRHGC wines were observed. In brief, vegetative attributes (green bell pepper-aroma and flavour) were statistically higher in the Winery 7 MonCabCort samples, suggesting that the characteristics of the two varieties (Monarch and Cabernet Cortis) were more noticeable. This sample was aged for 10 months in old oak barrels and qvevri, according to the label. In contrast, the liquorice-aroma (spicy) and woody flavour were slightly higher in MalFochCabCort samples from Winery 7, suggesting that this sample might be aged in oak barrels, even if no statistically significant differences were observed. According to the label, it had undergone contact with oak for 18 months. Tobacco-aroma and oak-aroma (woody) were slightly statistically significant in CabCort samples from Winery 3. Based on the standard winemaking methods used by this winery (Winery 3), this sample was aged for 24 months in American oak barriques. Indeed, oak aroma and tobacco aroma were also perceived by the panel.

Similarly to Moser et al.⁹¹, the wines studied did not have the ‘foxy’ off-aromas and flavours typical of some first-generation DRHGCs. Unlike Moser et al.⁹¹, however, other unusual sensory characters such as strawberry-cherry, cotton candy, and burnt-sugar candy were not noted by the panel.

Basic oenological parameters and CIELab colour

There were statistically significant differences (p ≤ 0.05) between the studied wines for free and total sulfur dioxide, tartaric acid, lactic acid, glucose and fructose, and alcohol per cent by volume, but not for pH. However, the Tukey’s Honestly Significant Difference (HSD) test showed that the V. vinifera wine was only significantly different (p ≤ 0.05) to the DRHGC wines for tartaric and lactic acid, with the V. vinifera wine having significantly higher levels of tartaric acid and lower levels of lactic acid than the DRHGC wines. The Tukey’s HSD test also showed that the V. vinifera and DRHGC wines did not differ significant terms of glucose and fructose, with one wine (Winery3_CabCort) having 0.87 g L⁻¹ of sugar and the others between 2.1 to 2.86 g L⁻¹. Similar results were seen in the MFA (Fig. 3c). DRHGCs have been noted as having higher levels of malic acid than V. vinifera, and this may explain the higher levels of lactic acid post-malolactic fermentation².

For CIELab colour, there were also statistically significant differences (p ≤ 0.05) between the studied wines for all CIELab parameters. However, the V. vinifera wine was not significantly different from the DRHGC wines, based on the Tukey’s HSD test.

Overall profile

The MFA plots (Fig. 3) show that the wine samples were separated into four groups. In detail, Fig. 3b shows that the first partial axis mainly separated the DRHGC wines, while the second axis differentiated the V. vinifera wines from the DRHGC wines. The first partial axis explained 34.6% of the data, the second explained 22.3%, and the third explained 12.0%.

Looking at the contributions of variable groups to Dim.1, it is clear that the volatile (GC×GC-MS) dataset made the greatest contribution, followed by mRATA and then by the LC-MS polyphenol dataset. In detail, the volatile (GC×GC-MS) dataset contributed 13.6% to the first axis, 14.8% to the second axis, and 18.3% to the third axis; mRATA sensory analysis contributed 13.0% to the first dimension, 5.39% to the second, and 9.24% to the third; LC-MS of polyphenols contributed 12.9% to the first dimension, 17.9% to the second, and 19.8% to the third; projective mapping sensory analysis contributed 12.3% to the first dimension, 7.42% to the second, and 15.5% to the third; CIELab parameters contributed 10.4% to the first dimension, 5.76% to the second, and 2.52% to the third; anthocyanins contributed 9.22% to the first dimension, 19.6% to the second, and 8.38% to the third; and PAC contributed 8.79% to the first dimension, 2.81% to the second, and 12.5% to the third (Fig. 3a).

The MFA Scores plot (Fig. 3b) shows separation of the DRHGC and V. vinifera wines along the second axis and the individual DRHGC wines on the first dimension. This shows that the analysis was able to separate the V. vinifera wine from the DRHGC wines based on the combined datasets (basic oenological parameters, CIELab, anthocyanins, proanthocyanidins, polyphenols, volatiles, and sensory datasets). This reinforces the distinctiveness of the DRHGC wines included in the study from the V. vinifera wine.

This study examined the chemical and sensory properties of both DRHGC and V. vinifera wines produced in South Tyrol, Italy. It shows that DRHGC wines are distinct from V. vinifera wines and that consumers do not inevitably prefer V. vinifera wines over DRHGC wines. This should be of interest to producers who are considering planting DRHGCs, and to wine appellation authorities who are considering authorisation of DRHGC within their appellations. It is also of interest given the requirements to reduce the use of pesticides, given the critical role that DRHGCs can play in this.

The DRHGC wines have unique c-PAC profiles. In brief, c-tetramer (m/z 1153) and c-pentamer-1 (m/z 1441-1) were higher in CabCort samples from Winery 3 and lower in MonCabCort and MalFochCabCort samples from Winery 7. In contrast, c-OH-tetramer (m/z 1169) was higher in MonCabCort and MalFochCabCort samples from Winery 3.

The PAC profile (conventional proanthocyanidins) showed interesting results. OH-trimers (m/z 867) showed the highest content in DRHGC samples from Winery 3 (MonCabCort and MalFochCabCort). In general, pentamer (m/z 1443) and OH-tetramer (m/z 1171) were higher in CabCort from Winery3 and CabSauv from Winery 8.

The volatile profile also differed significantly between DRHGC and V. vinifera wines. Winery 3 CabCort was defined by several key volatiles, including ethyl succinate, ethyl isopentyl succinate, benzeneacetaldehyde, and phenylethyl alcohol, while Winery 7 MonCabCort was defined by compounds including ethyl 2-hexenoate, nerol oxide, benzaldehyde, and α-terpineol.

These differences were also reflected in the sensory profile of the wines. In particular, the V. vinifera wine was warmer and more astringent than the DRHGC wines included in this study, and with more strawberry and jam aromas. In contrast, the DRHGC wines were higher in green bell pepper aroma and flavour (Winery 7 MonCabCort), liquorice aroma and woody flavour (Winery 7 MalFochCabCort), and tobacco and oak aromas (Winery 3 CabCort). Panellists preferred wines characterised by woody aromas and spicy notes. Bitterness and saltiness were also associated with overall quality.

This study characterised the chemical and sensory properties of several representative DRHGC and V. vinifera wines from South Tyrol, Italy, and determined some of the key features of red DRHGC wines from South Tyrol. The limited availability of red DRHGC wines in South Tyrol reflects the fact that this region is primarily a white wine-producing region. A forthcoming paper will examine white wines from the region.

Further studies could include wine from different regions, to further shed light on the distinctive nature of red DHRGC wines. In addition, consumer trials could be useful to determine consumer responses to these wines and to provide suggestions for future consumer education regarding the benefits of DRHGC wines. The usefulness of DRHGC in driving sustainable wine production methods should also be emphasised. Such developments could assist with the adoption of DRHGCs in South Tyrol and in other regions, which could have positive impacts on social, economic, and environmental sustainability.

Methods

Samples

Commercially available South Tyrolean red wines were obtained from local wineries, and care was taken to contrast DRHGC and V. vinifera cultivars with shared pedigree. Indeed, ‘Cabernet Cortis’ was selected as a DRHGC cultivar, and ‘Cabernet Sauvignon’ was selected as a conventional V. vinifera cultivar because ‘Cabernet Sauvignon’ is a parent cultivar for ‘Cabernet Cortis’. ‘Cabernet Cortis’ was bred from ‘Cabernet Sauvignon’ and ‘Solaris’ in 1982 by Nobert Becker in Freiberg, Baden-Würtemberg, Germany. It has good resistance to downy mildew and bunch rot and is very early ripening⁹².

Three wineries were included in the project. From each winery, a triplicate for each wine sample was collected.

Winery3_CabCort was harvested from a vineyard located in Kaltern-Caldaro, South Tyrol (BZ), Italy, at an altitude of 425 metres above sea level. This ‘Cabernet Cortis’ (100% monovarietal) from Winery3_CabCort is from the 2018 vintage. This particular cultivar is classified as a DRHGC.

Winery7_MonCabCort was harvested from a vineyard situated in Mühlbach-Rio di Pusteria, South Tyrol (BZ), Italy, at an altitude of 777 metres above sea level. It is a blend between ‘Monarch’ & ‘Cabernet Cortis’ (at a 20:80 ratio), both of which are DRHGC, from vintage 2021.

Winery7_MalFochCabCort was also harvested from a vineyard located in Mühlbach-Rio di Pusteria, South Tyrol (BZ), Italy, at 777 metres above sea level. It is a blend between ‘Maréchal Foch’ & ‘Cabernet Cortis’ (at a 20:80 ratio). The vintage year for this DRHGC wine is 2021.

Winery8_CabSauv was harvested from a vineyard in Kaltern-Caldaro, South Tyrol (BZ), Italy, at an altitude of 425 metres above sea level. Winery8 specialises in Cabernet Sauvignon (100%, monovarietal). This cultivar is classified as V. vinifera, and the wine is from vintage 2019.

Chemicals

Formic acid (LC-MS grade), acetonitrile (LC-MS grade), methanol (LC-MS grade) were purchased from Di Giovanni srl (Bologna, BO, Italy). Malvidin 3,5-O-diglucoside, sodium chloride anhydrous, 2-methyl-3-pentanol, and all food-grade products (caffeine, tartaric acid, lactic acid, malic acid) were purchased from Sigma–Aldrich (St. Louis, MO, USA). Food-grade 95% alcohol and sodium chloride were purchased from local markets, as were food-grade products that were not available from laboratory suppliers. Deionised water was produced using an Arium Mini Ultrapure Lab Water System (Santorius, Göttingen, NI, Germany).

Sensory analysis

Sensory analyses of DRHGC and V. vinifera wines were developed in three stages:

• (1) definition of attributes (visual, olfactory, and gustatory (taste and flavour)) by the panel (round table),

• (2) rapid training (2 sessions for each type of attribute, 1 h each), and

• (3) projective mapping (Napping ®) combined with mRATA (modified rate-all-that-apply) were applied during the sensory session (1 session, 1 h each).

Cysensy, an SQL binding sensory analysis web software developed in collaboration with the Computer Science faculty of the Free University of Bozen/Bolzano, was used during the training and the mRATA sessions.

Projective mapping was performed using GeoGebra Classic tool (https://www.geogebra.org/classic). To create the napping sheets, a rectangle was drawn on the GeoGebra sheet and its size was set as 52 × 35 cm (an example is provided in Supplementary Fig. S1), as reported by previous authors^93,94.

The panellists were recruited voluntarily and they were asked to sign an informed consent form in which they were informed about the presence of sulfite in the sample wines to avoid any related issues. The maximum quantity of wine per glass per session was 30 mL per glass per session⁵² and the wines contained approximately 13% v/v ethanol (3.9 g of ethanol per 30 mL sample). Three replicates each of four wines were tasted, split over two sessions. The study was conducted following the Declaration of Helsinki⁹⁵. There was no cost to participate; refusal to participate did not imply any penalties or loss of benefits, and participants were permitted to withdraw at any time without giving any reason. Based on this statement and since the participants were instructed to spit the wine after the analysis and the sample were commercial products available on the market, approval from an ethics committee was not required⁵². The panel was formed of twelve assessors, comprising 6 females and 6 males aged between 25–35 years old. In all three phases previously reported, water and unsalted crackers were provided for palate cleansing. The training and sensory sessions were conducted in the same room and at the same time for 1 month, to minimise the impact of location and time on the reproducibility of the sensory analysis.

First, the project and wines were introduced and presented to the panellists to give them insight into DRHGC and V. vinifera wines. During the round table session, panellists were instructed to individually assess each wine and to list descriptors related to its visual, olfactory, and gustatory characteristics using an interactive web whiteboard (Jamboard, Google, Mountain View, CA, USA). Following each sample, the panel leader and panellists engaged in a discussion to determine the most suitable shared attributes for each wine type, ultimately compiling the final sensory lexicon.

The panellists were subsequently trained for those specific attributes (Supplementary Data S1). Panellists were trained with reference standards⁸⁹ or natural standards if references were not available⁵². The definition of each sensory descriptor is given in Supplementary Data S1. The first session was specific to the aroma (olfactory) attributes, while the second session tested the gustatory attributes (taste and flavour) and the aroma attributes that had not been recognised (by more than 75%) during the first training session. The results of the training sessions are reported in Supplementary Data S2.

Following completion of the training and before starting the session, panellists were provided with detailed instructions on the definition of descriptors and how to conduct the sensory evaluation (sampling from the left to the right direction according to the order proposed) by the panel leader. They were reminded that the descriptors were related to olfactory evaluation (Supplementary Data S1), gustatory evaluation (Supplementary Data S1), and overall quality judgement (OQJ)⁹⁶. OQJ was defined as an overall evaluation of three aspects of the sensory analysis (visual, olfactory, and gustatory), and should be considered an objective evaluation of the overall quality of the product⁷⁰.

Each sample was evaluated by the panellists using a modified RATA (mRATA) and projective mapping methods. The bottles were opened 30 min before the session. Each sample was prepared in triplicate and a total of twelve 30 mL wine samples were assessed by each panellist across two different sessions (6 bottles per session) over two successive days. Wines were served in a randomised Latin Square incomplete block design to control for possible carry-over effects in each session⁹⁷. The wine samples were labelled with random 3-digit codes, and they were served in ISO glasses (ISO 3591:1977).

Regarding the mRATA, the panellists were asked to rate the intensity of each attribute from 0 to 5, where 0 was “does not apply”, 1 was “weak”, and 5 was “strong”. Additionally, panellists performed projective mapping by positioning each sample on the GeoGebra file (an example is reported in Supplementary Fig. S1) and the X and Y coordinates of each sample were collected to the nearest centimetre from the lower left corner of the rectangle in the GeoGebra sheet. This method was used to achieve a rapid categorisation and characterisation of wines.

Oenological parameters

Before all analyses, all wine samples underwent filtration using 0.22 µm syringe filters (Exacta+Optech Labcenter SpA, San Prospero, MO, Italy).

Basic oenological parameters were analysed on a MIURA One automatic analyser (Exacta+Optech Labcenter SpA, San Prospero, MO, Italy) for L-tartaric acid, L-malic acid, L-lactic acid, glucose and fructose, total sulfur dioxide, free sulfur dioxide, and total polyphenols (reagent kits purchased from Exacta+Optech Labcenter SpA, San Prospero, MO, Italy). Each parameter was calibrated against the relative reference standard provided by the supplier. The pH was measured using an XS pH 60 VioLab benchtop pH metre (XS Instruments, Carpi, MO, Italy) previously calibrated at pH 7.0 and 4.0⁷⁰. The results are reported in Supplementary Data S3.

CIELab colorimetry

CIELab measurements were made using a CR-400 colorimeter (Konica Minolta, Chiyoda-ku, Tokyo, Japan). Measurements were made in triplicate, and the means were used for analysis.

In CIELab, parameter L* indicates lightness (L* 0 = black, L* 100 = diffuse white), parameter a* indicates green-red colours (negative values indicate green, positive values indicate red), and parameter b* indicates blue-yellow colours (negative values indicate blue, positive values indicate yellow)^56,57,58.

Further derived parameters are h* and C*. Of these, h* indicates hue and is calculated from h = arctan b*/a*, while C* indicates chroma (saturation) and is calculated from [(a*)² + (b*)²]^0.556,58. The results are reported in Supplementary Data S3.

LC-MS analysis

Monomeric non-anthocyanin polyphenols were analysed using an LC-MS method adapted from a previous publication⁷⁰. The LC-MS analysis of polyphenols was performed using a LC/TQ 6465 UHPLC-QqQ/MS instrument (Agilent, Santa Clara, CA, USA) equipped with a 1260 Infinity II UHPLC with a quaternary pumps system, a 1260 Infinity II WR PDA detector, in series to an AJS ESI QqQ mass analyser. A Poroshell 120 SB-C18 2.1 mm × 100 mm × 2.7 µm column (Agilent) kept at 30 °C was used for separation of the compounds. The mobile phases were A, 0.1% v/v formic acid in Millipore water and B, 0.1% v/v formic acid in LC-MS analytical grade acetonitrile. The gradient separation programme was 1% mobile phase B from 0 to 1.5 min, 1 to 30% B from 1.5 to 19 min, 30 to 99% B from 19 to 20 min, 99% B from 20 to 24 min, 99 to 1% B from 24 to 25 min, and 1% B from 25 to 30 min. The injection volume was 5 µL. The PDA detector was set to record the absorbance in the 200–700 nm wavelength range using a 4 s response time (1.25 Hz) and 4 nm slit width, with 1 nm spectrum steps. The MS detection was performed in ESI- ionisation mode, with the following parameters: mass range, m/z 200–750; scan time, 500 ms; step size, 0.1 amu; fragmentor potential, 135 V; cell acceleration, 5 V; N₂ gas temperature, 340 °C; N₂ gas flow, 13 L min⁻¹; nebuliser pressure, 50 psi; sheath gas heater, 350 °C; sheath gas flow, 12 L/min; capillary voltage, −3500 V; nozzle voltage, −500 V. Tentative compound identification was carried out using full-scan mass spectrometry determination and the relative PDA λmax assignment to classify the compounds. Where complete identification could not be achieved, compounds were identified at least to a phenolic class. Targeted MS/MS fragmentation experiments (product ion monitoring, PRM) on selected ions were carried out with the following setup: source parameters as in the MS1 analysis; acquisition mass range from m/z 25 to m/z + 10 from the selected precursor ion; scan time, 125 ms; fragmentor potential, 135 V; collision energy, 25 V; and cell accelerator, 5 V. The results are reported in Supplementary Data S3.

Oligomeric proanthocyanidins (PAC) were analysed using an LC-MS method adapted from a previous study⁶⁵. Oligomeric proanthocyanidins (PAC) were separately analysed on the same UHPLC-QqQ/MS instrument (Agilent LC/TQ 6465 system). The chromatographic separation was carried out on a Vertex Plus Eurospher II (KNAUER, Berlin, Germany) column (4.6 mm × 250 mm × 5 µm) with a precolumn that was kept at 30 °C. The separation was carried out at a 0.7 mL min⁻¹ flow rate. The mobile phases were A, 0.1% [v/v] formic acid in Millipore water and B, 0.1% [v/v] formic acid in LC-MS analytical grade acetonitrile. The gradient separation programme was 1% mobile phase B from 0 to 2.5 min, 1 to 25% B from 2.5 to 50 min, 25 to 99% B from 50 to 51 min, 99% B from 51 to 55 min, 99 to 1% B from 55 to 56 min, and 1% B from 56 to 62 min. The injection volume was 5 µL. The mass spectrometer operated in ESI+ mode. Detection of the proanthocyanidins was performed in single-ion monitoring mode. In Supplementary Data S4 parameters for acquisition for all wines are listed. The parameters for mass spectrometric detection were as follows: fragmentor, 135; cell acceleration, 5 V; N₂ gas temperature, 230 °C; N₂ gas flow, 8 L min⁻¹; nebuliser pressure, 20 psi; sheath gas heater, 300 °C; sheath gas flow, 10 L min⁻¹; capillary voltage, 3000 V; and nozzle voltage, 2000 V. Feature selection (retention time/m/z ratio) was conducted by directly inspecting the acquired data, and selected features were used to directly extract the peak intensities in MzMine3⁹⁸ and create a peak table. These results, including the retention time of each compound and tentative assignations, were compared to those reported by Darnal et al.⁶⁵. The results are reported in Supplementary Data S3 and S5.

Anthocyanins were analysed using an LC-MS method that was adapted from De Rosso et al.⁹⁹. Anthocyanins were separately analysed on the same UHPLC-QqQ/MS instrument (Agilent LC/TQ 6465 system) used for polyphenols and PAC. The chromatographic separation was carried out instead on a Vertex Plus Eurospher II (KNAUER, Berlin, Germany) column (C18H, 4.6 mm × 250 mm × 5 µm) with a precolumn and kept at 30 °C, with a working pH range from pH 1. The separation was carried out at a 0.7 mL min⁻¹ flow rate. The mobile phase was A, 10% v/v formic acid in Millipore water, and B, 10% v/v formic acid in LC-MS analytical grade methanol. The gradient separation programme was 5% B from 0 to 3.5 min, 13.5% B from 3.5 to 50 min, 40% B from 50 to 60 min, 75% B from 60 to 61 min, 95% B from 61 to 66 min, 5% B from 66 to 67 min, and 5% B from 67 to 76 min with a flow rate of 0.7 mL min⁻¹. Injection volume was 5 µL. The PDA detector was set to record the absorbance in the 200–800 nm wavelength range using a 4 s response time (1.25 Hz) and 4 nm slit width, with 1 nm spectrum steps. The MS detection was performed in ESI+ ionisation mode, with the following parameters: mass range, m/z 250–1500; scan time, 500 ms; step size, 0.1 amu; fragmentor potential, 140 V; cell acceleration, 5 V; N₂ gas temperature, 230 °C; N₂ gas flow, 8 L min⁻¹; nebuliser pressure, 20 psi; sheath gas heater, 300 °C; sheath gas flow, 10 L min⁻¹; positive voltage, 3000 V; negative voltage, 2500 V; and nozzle voltage, −500 V. Tentative compound identification was carried out using full-scan mass spectrometry determination and the relative PDA λmax assignment to classify the compounds. Where complete identification could not be achieved, compounds were identified at least to a phenolic class. Results on targeted MS² fragmentation experiments (product ions monitoring—PRM) on selected ions were reported in Supplementary Data S3 and S6. Compound semi-quantification (mg L⁻¹) for DRHGCs and V. vinifera wines was obtained based on the calibration of the reference standard of malvidin-3,5-O-diglucoside (Mv3,5 G, [M]⁺, m/z 655.1; RT, 20.9) by linear regression (linear calibration range, 0.1–1 mg mL⁻¹; regression adj.-R², 0.9668; regression slope, 2E + 09x; and regression intercept set to 0)¹¹. The results are reported in Supplementary Data S7.

HS-SPME-GC×GC-ToF/MS analysis of volatile compounds

Volatile compounds were analysed using an HS-SPME-GC×GC-ToF/MS method applied that was slightly adapted from published studies^52,70. The samples were analysed at the same time, which corresponded to the time at which the bottles were opened, to minimise sampling error and oxidation. The sample preparation was carried as follows: 0.5 g of sodium chloride, 4 mL of wine sample, 10 µL of 2-methyl-3-pentanol as internal standard (from a stock of 1/50 of I.S. in ethanol) were placed into a 10 mL solid phase microextraction (SPME) glass vial. The vial was closed with a perforable screwcap with a silicon septum for automated SPME. SPME was performed using an LPAL3 GC autosampler (LECO Corporation, St Joseph, MI, USA) equipped with a Peltier stack, where the samples were kept at 4 °C before analysis. A Stableflex SPME fibre 50/30 µm DVB/CAR/PDMS with a 23 Ga-sized needle (Supelco, Bellefonte, PA, USA) was used. The parameters applied for the SPME sample extraction were:

• (1) sample incubation at 40 °C for 15 min with agitation at 300 rpm (alternating 5 s agitation and 2 s without agitation),

• (2) sample extraction for 30 min (fibre penetration depth in the vial, 25 mm), and

• (3) sample injection with 6 min desorption time in the GC split/splitless inlet (fibre penetration depth, 40 mm).

The SPME fibre was preconditioned at 240 °C for 6 min prior to analysis of each sample. The separation was performed by comprehensive GC×GC using an Agilent 7890B GC instrument (Agilent, Santa Clara, CA, USA) coupled with a Pegasus BT 4D time-of-flight mass spectrometer (LECO) and equipped with a Flux flow modulator (LECO). The separation was carried out at 1 mL min⁻¹ (He carrier gas) in splitless mode, with 2 mL min⁻¹ septum purge flow and 6 mL min⁻¹ inlet purge flow. The inlet temperature was 240 °C. The separation was carried in GC×GC mode on a polar cross-bond PEG-phase MEGA-Wax Spirit 0.30 µm × 0.18 mm × 40 m column (Mega S.r.l., Legnano, MI, Italy) as the first dimension and a Rxi-17 Sil 0.10 µm × 0.10 mm × (0.7 + 0.31) mm column (Restek, Bellefonte, PA, USA) as the second dimension. The two columns were connected by a Flux modulator (LECO), with a 15.92 psi auxiliary diverting helium flow. The second column was mounted in a secondary oven that is inside but thermally isolated from the larger primary oven. The primary oven temperature programme was 40 °C for 6 min, 40 to 180 °C at 3 °C min⁻¹, 180 to 240 °C at 10 °C min⁻¹, and 240 °C for 1 min. The secondary oven had a temperature 5 °C higher than the first oven, and its temperature programme was changed in synchrony with the main oven to maintain this temperature difference. The GC×GC instrument was connected to the mass spectrometer by a transfer line kept at 250 °C. The mass spectrometer operated with the following parameters: 0 s acquisition delay, 70 eV filament electron energy, 250 °C ion source temperature, 150 spectra/s acquisition rate, 32 kHz extraction frequency, and an acquisition mass range of m/z 35–530. The mass spectrometer was tuned each time before starting a new sequence of analyses. After each analysis, each GC×GC chromatogram was reconstructed as 2D colour maps by deconvolution using ChromaTOF® (LECO) ver.2021 software. Automatic compound assignment was carried out using NIST 2017 (NIST MS Search 2.3) database comparison. Linear retention indexes were calculated from the first-dimension retention times against the injected series of linear fatty ethyl esters C4-C24 (even carbon saturated, Merk, Darmstadt, Hesse, Germany). The collected dataset underwent automatic alignment using ChromaTOF® (LECO) ver. 2021 processing software. The obtained aligned table was inspected and refined prior to the statistical analysis. The results are reported in Supplementary Data S3 and S7.

Statistical analysis

Statistical analysis was performed using GNU R software version 4.3.3¹⁰⁰ and ‘tidyverse’ version 2.0.0¹⁰¹. MFA was applied using the ‘MFA’ functions of the ‘FactoMineR’¹⁰² and ‘factoextra’¹⁰³ packages. Individual PCAs were automatically performed by the ‘MFA’ function (nested within the ‘separate analyses’ sub-list of the MFA object) and were extracted and graphed separately.

To determine whether differences between DRHGC and V. vinifera wine samples were statistically significant, analysis of variance (ANOVA) was conducted to determine significance (1-way ANOVA, p ≤ 0.05) using the base R ‘aov’ function from the ‘stats’ package¹⁰⁰. Tukey’s HSD analysis was also performed for multiple comparisons by using the ‘TukeyHSD’ function (α = 0.05) from the ‘stats’ package¹⁰⁰ or the ‘HSD.test’ function from the ‘agricolae’ package¹⁰⁴. Significance was represented by groups and indicated by alphabetical letters, determined using the ‘HSD.test’ function from the ‘agricolae’ package. Stars indicating significance levels were determined using ‘stars.pval’ function from the ‘gtools’ package¹⁰⁵. The results are reported in Supplementary Data S9.

Each variable was investigated to prepare a complete dataset for statistical analysis, and the missing values were separately handled in each dataset via the following process:

• (1) when NAs were present in at least two replicates in a wine sample, nonlinear Iterative Partial Least Squares (NIPALS)¹⁰⁶, which was built to handle NAs, was used to determine missing values using the XLSTAT add-on for Excel (Lumivero, Denver, CO, USA) (estimated numbers in black bold, Supplementary Data S3);

• (2) if the NAs were present in all replicates in a wine sample, the NAs were handled by replacing them with random small values between zero and the LOD¹⁰⁷ (estimated numbers in blue bold, Supplementary Data S3).