TEST3 OENO One Sensory assessment of grape polyphenolic fractions: an insight into the effects of anthocyanins on in-mouth perceptions 1 Dipartimento di Scienze Agrarie, Forestali e Alimentari. Università degli Studi di Torino, Grugliasco, Italy 2 ISVV, EA 4577 Oenologie, F-33140, Université de Bordeaux, Villenave d’Ornon, France 3 Inrae, ISVV, USC 1366 Œnologie, 33140, Villenave d’Ornon, France 4 Wine School, Ningxia University, Yinchuan, Ningxia, 750021, P.R. China Paissoni MariaAlessandra 1 2 3 * Waffo-Teguo Pierre 2 3 Ma Wen 2 3 4 Jourdes Michael 2 3 Giacosa Simone 1 Segade SusanaRío 1 Teissedre LucaRolle1andPierre-Louis 2 3 * Keywords: anthocyanins, winegrapes, check-all-that-apply (CATA), descriptive analysis (DA), sensory analysis, polyphenols Introduction

In red wines, colour and “in-mouth” features strongly influence the sensory perception of quality (Peynaud, 1987; Parpinello et al., 2009, Piombino et al., 2020). These characteristics are mainly connected to the polyphenols extracted from black grapes. Among them, two groups are particularly relevant in terms of content: anthocyanins and condensed tannins (also generally referred to as proanthocyanidins), and their constituting monomers, the flavan-3-ols. Anthocyanins are responsible for the red wine colour, whereas condensed tannins are involved in both colour stabilisation and the “in-mouth” characteristics of wine, such as astringency and bitterness (Ma et al., 2014). Recently, wine research has focused on the improvement of colour parameters and the modification of “in mouth” features with the development of technologies and practices aiming to improve the initial grape phenolic content and composition, their extraction into grape juice during maceration, and wine ageing management (Harrison, 2018). Nevertheless, wine tannin concentration and properties alone do not represent the full “in-mouth” complexity of wine. Wine technological parameters, such as ethanol content, total acidity and pH influence perception of astringency (Pickering and Demiglio, 2008; Fontoin et al., 2008, Laguna et al., 2017). Furthermore, in wine, other macromolecules (e.g., polysaccharides, proteins, and ellagitannins) derived from grape, yeasts, or external sources (e.g., wood used in ageing) can modulate the tannin effect or directly elicit astringency sensations (Glabasnia and Hofmann, 2006; Fukui et al., 2002; Laguna et al., 2017).

The role of anthocyanins in particular has been investigated to better understand how and to what extent they can modulate “in-mouth” properties (Table 1). Although they are known to contribute to depleting the astringency of tannins with the formation of complexes (Vidal et al., 2004a) by the reduction of the available -OH group interacting with salivary proteins, the direct eliciting capacity of anthocyanins is far from clear. Pure anthocyanins have been reported to have a very “mild indistinctive taste” (Singleton and Noble, 1976), supporting the hypothesis that their role may only be relevant to the pigments formed with flavan-3-ols. In contrast, the addition of a mixture of isolated grape anthocyanins to seed and skin extracts has been found to increase astringency compared to the unspiked fractions (Broussard et al., 2001). These results have been supported by winemaking experiments in which anthocyanins or grape pomace were added to white grape juices before fermentation, showing that anthocyanins increase astringency sub-qualities related to surface smoothness (fine grain), as well as other “in-mouth” attributes, such as “dry”, “grippy” and perceived viscosity in final wines (Oberholster et al., 2009). These descriptors agreed in part with the results of previous studies, in which anthocyanins were tasted in model wine solutions: they were found to contribute to “fullness”, as well as to “chalkiness” and “coarseness” sensations (Vidal et al., 2004b), and to increase “dryness” and “roughness” (Vidal et al., 2004c). Anthocyanin purification of grape skins and wines in high quantities is still a challenge, since there are several compounds which interfer with sensory properties (mainly flavonol-belonging molecules) and which are difficult to avoid in extraction steps. In an improvement in purification, fractions of free glucoside and p-coumaroylated anthocyanins were shown not to differ from the unspiked model wine solution in both the wine range (13 % v/v) and at a reduced ethanol level (5 % v/v) (Vidal et al., 2004a). Besides fraction composition and purity, such different results from sensory analyses may be due to the variability of the sensory technique used (e.g., different descriptors, reference standards and matrix solutions), as well as to differences in assessor training (Gawel, 1997; Gibbins and Carpenter, 2013; Sáenz-Navajas et al., 2016).

Another possible option for evaluating the sensorial impact of polyphenols is the instrumental analysis of astringency and bitterness (Table 1). Astringency is a complex sensation involving several mechanisms, from sensory active molecules interacting with different salivary protein families to the formation of soluble and precipitable complexes. The onset of astringency involves hydrogen bonds, as well as hydrophobic and electrostatic interactions between sensory active molecules and proteins, and their complexity increases in mixed polyphenol solutions (Soares et al., 2019). Moreover, additional phenomena can occur, such as the direct activation of mechanoreceptors, or direct interaction with the mouth epithelium (Gibbins and Carpenter, 2013; Soares et al., 2016). Recently, the involvement of anthocyanins in the onset and formation of complexes has been demonstrated (Ferrer-Gallego et al., 2015; Paissoni et al., 2018; Soares et al., 2019). Considering bitterness, anthocyanins have been found to activate the corresponding taste receptors (Soares et al., 2013). Nevertheless, testing this stimulus in an alcoholic media is still impossible and it is difficult to study complex polyphenol matrices, as is the case for wine. Therefore, despite the possibility to instrumentally study the different mechanisms driving astringency sensations, the final perception is influenced by the complexity of wine polyphenols and other sensory active molecules mixture. Therefore, together with the instrumental evidence of anthocyanin involvement in astringency-driver mechanisms and bitterness stimuli, sensory analysis is still necessary for understanding the final in-mouth features solutions.

The aim of this study was to investigate the sensory properties of grape extracts, with a particular focus on grape anthocyanins. First, grape extracts were tasted to establish their sensory active thresholds, and the appropriate terminology was selected in order to evaluate their contribution to in-mouth sensory properties by using a check-all-that-apply (CATA) methodology. After that, the assessors received training in the selected categories and descriptors, and then carried out a descriptive analysis (DA) of different individual grape extracts: polyphenol extracts from skins and seeds and anthocyanin extracts divided according to acylation group. Moreover, to find out if anthocyanins can influence the in-mouth perception of wine, the anthocyanin extract and its derived fractions were added to the polyphenol-based extracts and evaluated following the same descriptive sensory procedure.

Table 1. Studies which have carried out sensory and instrumental analyses on anthocyanin sensory properties.
titre du tableau
Sensory analysis
Year Astringency Bitterness Mouthfeel Notes References
1976 "indistinct mild taste" Singleton and Noble (1976)
2001 + = Anthocyanin extract added to seed and skin tannins in white wine Brossaud et al.(2001)
2004 + dryness, roughness, chalkiness Anthocyanin extract (0.5 g/L, purity 78 %) added to model wine at 13 % ethanol Vidal et al. (2004c)
2004 + = fullness, chalkiness, coarseness Anthocyanin extract (0.5 g/L, purity 70 %) added to model wine at 11, 13, and 15 % ethanol Vidal et al. (2004b)
2004 in-mouth dryness Anthocyanin glucosides (0.5 g/L, purity 98 %) and coumaroylated (0.5 g/L, purity 87 %) in model wine at 13 % ethanol Vidal et al. (2004a)
= Anthocyanin glucosides and coumaroylated (0.5 g/L) in un-buffered, 5 % ethanol solution
2009 + = fine grain, dryness, grippiness, viscosity Anthocyanin extracts (purity 95 %) added at 1.44 g/L in white juice during winemaking Oberholster et al. (2009)
2014 + + Two p-coumaroylated anthocyanin contributors of bitterness, p-coumaroylated petunidin and malvidin-3-O-gucoside contributors to astringency Gonzalo-Diago et al. (2014)
2015 + Glucoside anthocyanin extract (purity 95 %), moderate velvety and astringency intensity Ferrer-Gallego et al. (2015)
2017 + + dryness, bitterness, persistency Oligomeric anthocyanins from wine fractions redissolved in 7 % ethanol buffer Sáenz-Navajas et al. (2017)
2018 + + Anthocyanin extracts, glucoside, acetylated, and coumaroylated fractions (purity 95, 98, 87 and 91 % respectively) in model wine at 12 % ethanol; BETa 255, 297, 68, and 58 mg/L, respectively Paissoni et al. (2018)
Chemical analysis
Year Astringency Bitterness Interaction notes
2013 + TAS2Rsb activation Malvidin-3-O-glucoside activates TAS2R7 bitterness receptors (threshold of activation 6.0 μM), whereas cyanidin-3-O-glucoside did not activate TAS2Rs studied Soares et al. (2013)
2015 + Precipitable/soluble complexes with saliva PRPsc Individual glucoside formed complexes with acidic PRPs. Evidence of soluble complexes between malvidin-3-O-glucoside with Histatin and PRPs. Ferrer-Gallego et al. (2015)
2018 + Precipitable complexes with saliva Decrease in acetylated and coumaroylated fractions when saliva was added. Precipitation occurs depending on the anthocyanin acylation. Paissoni et al. (2018)
2019 + Electrostatic interactions Malvidin-3-O-glucoside and (-) epicatechin synergic interaction with PRPs. The first driven by electrostatic interactions, the latter by hydrophobic and hydrophilic interactions. Interactions are increased in the presence of mixture. At high concentration non-specific reactions Soares et al. (2019)
2020 + Oral cell interactions Anthocyanins (glucosides extract) are retained by different oral cells (HSC-3 from tongue and TR146 from buccal mucosa). Saliva and mucosal pellicle on the cell line decreased the anthocyanins retained. Soares et al. (2020)

Symbols “+” and “=” indicate increased or unchanged perception respectively. a BET=Best estimated threshold, b TAS2Rs = human bitter taste receptors encoded by the TASTE 2 Receptor (TAS2R) gene family, c PRPs = Proline-rich proteins.

Materials and methods 1. Extraction and purification of grape polyphenols

Grape samples from Vitis vinifera L. cv Nebbiolo and Barbera were collected at maturity in the Piedmont region (Northwestern Italy) during the 2015 vintage. Grape skins were peeled using a laboratory spatula and grape seeds were manually removed. The obtained grape material was lyophilised and then ground in a ball grinder; the resulting powder was used to carry out the extraction of grape polyphenol fractions.

1.1. Anthocyanin extracts

Total anthocyanin (TA) and its derived glucoside (GF), acetylglucoside (AF), and coumaroylglucoside (CF) fractions were extracted and purified by means of Centrifugal Partition Chromatography (CPC) and preparative HPLC as described in detail by Paissoni et al. (2018. The final extract purity was 95 %, 98 %, 87 %, and 91 % for TA, GF, AF, and CF respectively, calculated using the area ratios of 520 nm and 280 nm by HPLC-DAD. Total anthocyanin extracts (TA) from Barbera and Nebbiolo were individually tasted in a preliminary evaluation (triangle test; ISO 4120:1983), whereas subsequent evaluations were performed on Barbera TA only. The fractions differentiated by acylation were obtained from a mixture of the two different varieties.

1.2. Total polyphenol extract

Skin and seed polyphenol extraction was performed as described by González-Centeno et al. (2012) with an ASE 350 Accelerated Solvent Extraction System (Dionex Corporation. Sunnyvale, CA, USA). A mixture of Barbera and Nebbiolo ground grape seeds and skins was used to produce the polyphenol extracts. Grape skins and seeds were separately subjected to eight solid/liquid consecutive extractions with acetone/water (80:20, v/v) for the solvent system (40 mL of the corresponding solvent system). The ASE experimental settings were reported in Ma et al. (2016). The extract was then evaporated under reduced pressure and lyophilised to create crude skin and seed extracts. The crude polyphenol extract (equal to 5 g of dried powder of skins or seeds) was solubilised in 250 mL of water/ethanol (95:5, v/v) and extracted three times with chloroform to remove the lipophilic material (Lorrain et al., 2011). This aqueous extract was concentrated and lyophilised to obtain a dry powder, indicated as SkTOT and SdTOT for skin and seed extracts respectively. The obtained extracts were analysed via phloroglucinolysis (Lorrain et al., 2011). Mean degree of polymerisation (mDP) was 15.5 ± 0.45 and 3.7 0± 0.10 for SkTOT and SdTOT respectively. The percentage of galloylation (G %) was 15.20 ± 0.27 for SdTOT and 2.20 ± 0.10 for SkTOT. The percentage of prodelphinidins (Pd %) was 34.00 ± 0.23 for SkTOT, whereas no prodelphinidins were detected in SdTOT. For the skin extract, anthocyanin content was determined by HPLC-DAD (Paissoni et al., 2018), accounting for 130.2 ± 2.8 mg malvidin-3-O-glucoside /g of extract, (Sigma–Aldrich, Saint Quentin Fallavier, France). All the extracts were lyophilised twice before sensory analysis to ensure the absence of solvents.

2. Sensory analysis of grape extracts 2.1. General procedure

The sensory analyses were conducted in a tasting room at the University of Bordeaux, Oenology research unit (ISVV, France). The room fulfilled the ISO 8589:2007 standard (sound insulation, constantly regulated temperature).

A panel of 18 volunteer assessors from the Oenology department at the University of Bordeaux (ISVV, France) took part in the experiment. All assessors are considered wine experts according to the definition of Parr et al. (2002). Nevertheless, a preliminary panel selection process was carried out, in which the assessors tasted standard solutions: aluminium sulphate (AlSu; 2000 mg/L) for astringency, quinine sulphate (Qsu; 15 mg/L) for bitterness, tartaric acid (5 g/L) for acidity, and catechin (1 g/L) for both astringency and bitterness (Chira et al., 2012). In order to do this, two tests were performed: a discrimination test (triangle test; ISO 4120:1983) and an identification test. In the triangle test, the assessors were asked to recognise different samples in the series composed of spiked deionised water solution and water without the standard molecules. For the identification test, the assessors were asked to taste four spiked water solutions, and to identify and describe the in-mouth sensations perceived. Only those who recognised the sensations elicited by the reference standards were accepted as assessors for the experiment (16 assessors remained in the experiment).

The assessors were informed of the purification methodology for obtaining the fractions conducted in the laboratory. If they chose to continue in the experiment, they signed a consent form. A total of nine sessions were conducted prior to the formal evaluation, comprising three preliminary assessments of the extracts and six training sessions on the attribute scales of the descriptive analysis (DA). Finally, the formal descriptive analysis (DA) of the investigated extracts was conducted in five sessions.

2.2. Preliminary extracts’ evaluation

2.2.1. Triangle test

Two sessions were dedicated to determining a suitable extract concentration for the experiment. Triangle tests (ISO 4120:1983) were conducted taking into account the usual wine concentration ranges and detection thresholds. Two total anthocyanin extracts (from different grape cultivars, Nebbiolo and Barbera TA; 250 mg/L), glucoside fractions (GF; 300 and 400 mg/L), acetylglucosides and coumaroylglucosides fractions (AF and CF; 100 mg/L each), skin polyphenol extract (SkTOT; 500 mg/L), and seed polyphenol extracts (SdTOT; 500 and 750 mg/L) were tasted. The selected extracts were dissolved in model wine solutions (12.5 % v/v ethanol, 4 g/L of tartaric acid, pH 3.5), and were randomly tasted along with the unspiked model wine solutions. Nebbiolo vs Barbera TA extracts were also tested to find any differences between the two. Barbera TA vs un-spiked model wine and Barbera TA vs Nebbiolo TA were evaluated in duplicate (once per session) to confirm the obtained results. Seed extract and skin extract at 500 mg/L, GF at 300 mg/L, CF and AF at 100 mg/L were evaluated in the first session; whereas seeds extract at 750 mg/L and GF at 400 mg/L were evaluated in the second session, with the aim of achieving higher discrimination from the unspiked model wine. In addition to the triangle tests, the assessors were asked to report the descriptors that helped them identify the correct sample.

2.2.2. Check-All-That-apply (CATA) for the selection of descriptors

Four extracts representative of the groups under evaluation were selected for a Check-All-That-Apply (CATA) analysis, in order to identify the most frequently reported attributes and the terms which were able to best discriminate the extracts (Varela and Ares, 2012). The selected grape extracts were skin extract (SkTOT; 1000 mg/L), seed extract (SdTOT; 1000 mg/L), total anthocyanin extract from Barbera (TA; 400 mg/L), and glucoside fraction of anthocyanins (GF; 400 mg/L) in model wine solution (12.5 % v/v ethanol, pH 3.5, 4 g/L of tartaric acid). The selection of attributes was based on the following: results previously obtained by Paissoni et al. (2018), the descriptors reported in the triangle tests (only the ones by assessors who correctly discriminated among samples), and bibliographic research on the tasting of grape/wine extracts and fractions, in particular anthocyanins (Oberholster et. al., 2009; Gawel et al., 2000; Vidal et al., 2004a; Vidal et al., 2004b; Vidal et al., 2004c; Sáenz-Navajas et al., 2017). The selected attributes (23 in total) belonged to taste (sweet, bitter, salty, acid), and the selection of astringent sub-qualities was based on the mouth-feel wheel terminology proposed by Gawel et al. (2000). The introduced sub-qualities were: weight (watery, viscous, dense), texture (oily), heat (hot), irritation (prickle, tingle), drying (dry), and those related to astringency including surface smoothness (emery, silky) and particulates (dusty, grainy, chalky), complex terms (soft, round, mouthcoating, aggressive), a dynamic attribute (adhesive), and the term “rich”.

2.3. Descriptive analysis (DA)

2.3.1 Sensory training for descriptive analysis (DA)

The assessor training on astringency and bitterness scales was a slightly modified version of that proposed by Chira et al. (2012). During the first part (two sessions), the assessors were familiarised with the different sensations, using a stimuli concentration range. In the first session, low (500 mg/L AlSu and 2.5 mg/L QSu) and high (4000 mg/L AlSu and 30 mg/L QSu) reference standards of astringency and bitterness dissolved in water were evaluated. To improve the assessors' ability to distinguish bitterness from astringency, they were provided with solutions of the mixed stimuli and asked to score low/high astringency or bitterness; one solution was representative of low bitterness and high astringency (2.5 mg/L Qs and 4000 mg/L AlSu), and the other of high bitterness and low astringency (30 mg/L QSu and 500 mg/L AlSu). The assessors' scores allowed us to slightly modify the scale used for training based on the tasted fractions (section 2.2.2). Part of the second session focused on the reference standard solutions for astringency and bitterness in model wine solutions (12.5 % (v/v) ethanol, pH 3.5, 4 g/L of tartaric acid), by carrying out an ordination test of selected scales; i.e., for astringency (model wine, and model wine plus 1000, 2000, 4000, and 6000 mg/L AlSu) and for bitterness (model wine, and model wine plus 2.5, 5, 10, and 20 mg/L QSu). The assessors were also asked to distinguish in low, moderate, and high astringency and bitterness in mixed solutions (2.5, 10, 20 mg/L of QSu for bitterness; 1000, 4000, and 6000 mg/L of AlSu for astringency).

In the second part of the training (three sessions), the assessors were familiarised with the scales to be used for overall astringency and bitterness, and with the definitions of astringency sub-qualities. The assessors discussed the definitions of particulate (in-mouth sensation), surface smoothness (after expectoration) and irritation. Tactile standards were used when possible, whereas terms that did not have standards were agreed upon by written definition. Attribute rating was done on an 8-point scale: 0 = “absence”, 1 = “low” and 7 = “high”). “Low” and “high” levels were anchored with the use of reference standards for each attribute (Figure 1).

For example, for overall astringency, scores of 1,4 (corresponding to the average value) and 7 corresponded to 1000, 4000, and 6000 mg/L AlSu respectively, whereas for bitterness, these scores corresponded to 2.5, 10 and 20 mg/L QSu. Fabric samples were selected to anchor the surface smoothness line scale: “silky” (silk fabric) and “hard emery” (corduroy) were located along the scale at 1 and 7 points respectively. For the particulate attribute, “chalky” (talc powder) and “grainy” (white sugar) corresponded to 1 and 7 points. For irritation, definitions were given and placed on a low, moderate, and high scale (defined as “tingle”, “prickle”, and “hot” at 3, 5 and 7 respectively). The 0 score represented the model wine solution.

Figure 1. Tasting scorecard with anchored standard references.

Figure 1. Tasting scorecard with anchored standard references.

A summary of definitions, standard scales and references is provided in Table 2. During the last training session (the sixth), the assessors took part in a simulation of a descriptive analysis to become familiar with the tasting procedure and scorecard. They tasted solutions for stimuli enhanced by the samples under evaluation; in particular, a seed tannin extract from cv Nebbiolo (1250 mg/L), skin tannins (1000 mg/L skin commercial oenological tannin) and catechin and gallic acid solutions (both 1000 mg/L). At the end of each training session, the selected descriptors or intensity scores for each solution were compared in a discussion led by the panel leader.

Table 2. Descriptors, definitions, and reference standards used in descriptive analysis.
titre du tableau
Descriptor Standard or definition Reference
Bitterness (oral) Quinine sulphate (max 20 mg/L, average 10 mg/L)
Overall astringency (oral) Aluminium sulphate (max 6 g/L, average 4 g/L)
Particulate powder (manual) Feeling of particulate matter brushing against the surface of the mouth through the movement of the solution Gawel et al. (2001)
Talc: talc powder
Dusty: white flour
Sandy: grounded pepper
Grainy: white sugar
Surface smoothnessfabric (manual) Texture felt on mouth surfaces when the different surfaces came in contact with each other (after expectoration) Gawel et al. (2001)
Silky: silk
Velvety: velvet
Fine grain: fine cotton
Hard grain: corduroy
Irritation definition Sense of irritation usually associated with carbonation Pickering and Demiglio (2008)
Tingly (low)
Prickly (moderate)
Hot (high)

2.3.2. Formal DA of grape extracts

The assessors evaluated a maximum of four extract solutions per session for a total of five sessions, in each session one sample was repeated to evaluate the assessor’s performance (a maximum total of five samples per session). Each extract solution was coded with a three-digit random code, and black ISO 3591:1977 wine glasses were used and randomly placed in order not to bias perception.

In each evaluation session, the assessors were asked to assess the reference model wine solution (0 on the scale) and the medium point of the scale on which they were trained for overall astringency and bitterness (i.e., 4000 mg/L aluminium sulphate and 10 mg/L quinine sulphate for astringency and bitterness respectively) to help standardise the use of the scale before sample evaluation. Fabric samples for surface smoothness and powdery standards for particulates were provided during each session in each individual booth. The complete scales for bitterness and overall astringency were available in case assessors wanted to try them again before tasting. All fractions were served at room temperature and evaluated in individual booths. All extracts were evaluated in model wine (12.5 % v/v ethanol, 4 g/L tartaric acid, pH 3.5). The final concentrations were chosen according to the wine concentration range and triangular tests as explained above, which were 1000 mg/L for both SdTOT and SkTOT, 400 mg/L for TA and GF, and 100 mg/L for CF and AF.

The individual fractions of anthocyanins, TA, GF, and CF, were also added to the SkTOT and SdTOT solutions (1000 mg/L each) in the same concentration at which they were tasted individually (400 mg/L for TA and GF, and 100 mg/L for CF) to investigate if they could modify the sensory properties of other polyphenols in the wine concentration range.

To minimise fatigue and standardise the assessment process, a tasting and rinsing procedure was established: the assessors were asked to (1) sip and swirl the solution in their mouth for 5-10 s, (2) expectorate, and (3) fill out the scorecard. After each sample, the assessors were asked to rinse their mouth out with water, eat a piece of unsalted cracker and rinse again with water.

 3. Data analysis

Data were analysed using R software (R Core team, Version 3.5.0; R Studio, Version 1.1.453). The panel performance for CATA questions was checked using the reproducibility index (Ri), as described by Campo et al. (2008), and calculated on the replicate sample to check assessor repeatability. For the CATA descriptors of selected data, the attributes on the list were ranked according to their citation frequency to identify the most relevant descriptors for tasting and to better distinguish each extract. A Correspondence Analysis (CA) was performed on the CATA frequencies table (descriptors in rows and evaluated extract in columns) using Factominer R package (Lê et al., 2008). CATA significant attributes were assessed with Cochran's Q test (Varela and Ares, 2012). Since CATA was performed as a preliminary approach to underlining attributes potentially relevant for further investigation, the criteria for significance was increased to p < 0.1 to avoid missing relevant terms.

The panel performance for descriptive analysis (DA) was evaluated using PanelCheck software (version 1.4.2). The panel performance was tested on the subset of replicate samples by three-way ANOVA with “replicate”, “sample” (extract), and “assessors” as factors and their interaction. The evaluation was considered adequate when there were no significant differences (p > 0.05) in “assessors”*“replicates” and “assessors”*“sample” interactions. Principal component analysis (PCA) as Tucker-1 plots were performed for both attributes and assessors as described by Tomic et al., 2010. The assessors' projections were checked in the loading plot for bitterness, overall astringency, particulate, irritation and surface smoothness in order to assess their agreement on each attribute.

Two-way ANOVA (assessors as random and extracts as fixed factors; R package nlme, Pinheiro et al., 2020) was used to compare DA data from the formal sessions, and in case of significant differences, the Tukey HSD post-hoc test (R package multcomp, Hothorn et al., 2008) was performed to establish differences among extracts.

 Results and discussion 1. Triangle test results

The first experiment aimed to assess the concentrations at which the fractions could be perceived by the panel in a triangle test: the tested concentrations and results are reported in Table 3. The initial concentration to be tested was 500 mg/L total polyphenols from the skin (SkTOT) and seed (SdTOT) extracts. Although SdTOT had already been perceived as different by most of the panel (p < 0.05), SkTOT was not easily recognised as being different (p > 0.05). When the concentration of SdTOT increased to 750 mg/L, more differences were perceived (p < 0.001). Because the polyphenol range of a red wine can vary by up to several grams per liter, the concentration chosen for the tasting of these two fractions was 1000 mg/L, which fulfilled the red wine range and allowed the fraction to be clearly perceived by the assessors.

Concerning total anthocyanin extract (TA), two different extracts from Vitis vinifera L. cv Nebbiolo and Barbera (with a different ratio of individual anthocyanins and acylation; Paissoni et al., 2018) were tasted at a concentration of 250 mg/L. Both the extracts were perceived as being different to the unspiked model wine solution (p < 0.01). Interestingly, when these two fractions were compared (400 mg/L), they were not perceived as being different. Therefore, just one (TA from cv Barbera) was used for the following experiments. To improve sensitivity, a concentration of 400 mg/L was chosen, which is in accordance with young red wine concentrations and is similar to other sensory assessments (Vidal et al., 2004a; 2004b; Ferrer-Gallego et al., 2015).

Regarding the individual fractions, glucoside anthocyanins (GF) were perceived as different at 300 mg/L (p < 0.05), but more so at 400 mg/L (p < 0.01); thus the latter concentration was chosen for further analysis. Acetylated (AF) and p-coumaroylated (CF) anthocyanins were found to differ at 100 mg/L (p < 0.01 and p < 0.05 for AF and CF respectively). Therefore, these concentrations were chosen after the triangle tests, as they could be easily perceived by the assessors, although they are very high compared to the concentrations found in commercial wines.

Table 3. Triangle test results for the perception of phenolic fraction additions to the model wine solution.
titre du tableau
Sample Concentration(mg/L) Reference solution α value (correct answer)n = 16
Total anthocyanin extract (Nebbiolo, TA) 250 model wine 0.01 (11)a
Total anthocyanin extract (Barbera, TA) 250 model wine 0.01 (11)a 0.01 (11)b
Glucoside fraction (GF) 300 model wine 0.05 (9)a
Glucoside fraction (GF) 400 model wine 0.01 (11)b
Acetylated fraction (AF) 100 model wine 0.01 (11)a
Coumaroylated fraction (CF) 100 model wine 0.05 (10)a
Skin polyphenols (SkTOT) 500 model wine ns (7)a
Seed polyphenols (SdTOT) 500 model wine 0.05 (10)a
Seed polyphenols (SdTOT) 750 model wine 0.001 (12)b
Total anthocyanin extract (Nebbiolo, TA) 400 total anthocyanin extract (Barbera, TA) ns (8)a ns (4)b

a First session results and b and second session results.

2. CATA results

Assessor performance in a wine CATA analysis is considered to be adequate if Ri > 0.20 (Campo et al., 2008). In our case, assessors who failed this repeatability test were excluded from further analysis (n = 2). The overall panel performance (n = 10) was Ri = 0.43, which fulfilled the repeatability requirement. Only descriptors cited by at least 20 % of the panel were considered for further analysis. A frequency table summarises each descriptor individually, which were grouped into arbitrarily chosen categories based on bibliography (Gawel et al., 2000), namely taste, mouthfeel, astringency sub-qualities, and complexive astringency (Table 4).

Table 4. Frequencies of CATA analysis.
titre du tableau
Descriptor Grouping Frequencies of citation
Group Sub-qualitiesa Rank % in the group % in total p-valueb
Bitter (amer) Taste 1 46.9 13.3 0.464
Acid (acide) Taste 4 21.9 6.2 0.476
Sweet (sucré) Taste 13 7.8 2.2 0.738
Salty (salé) Taste 3 23.4 6.6 0.045
Hot (brûlant) Mouthfeel heat 4 22.2 6.2 0.651
Dry (asséchant) Mouthfeel drying 12 9.5 2.7 0.037
Tingly (piquant) Mouthfeel irritation 8 15.9 4.4 0.641
Prickly (pointu) Mouthfeel irritation 11 11.1 3.1 0.478
Oily (onctueux) Mouthfeel texture 7 17.5 4.9 0.069
Watery (acqueux) Mouthfeel weight 9 14.3 4.0 0.762
Dense (dense) Mouthfeel weight 12 9.5 2.7 1.000
Emery (rugueux) Astringency sub-qualities surface smoothness 6 24.0 5.3 0.383
Dusty (poussiéreux) Astringency sub-qualities particulate 5 26.0 5.8 0.697
Grainy (granuleux) Astringency sub-qualities particulate 12 12.0 2.7 0.242
Chalky (talc) Astringency sub-qualities particulate 8 20.0 4.4 0.742
Silky (soyeux) Astringency sub-qualities surface smoothness 9 18.0 4.0 0.040
Rich (gras) Complexive astringency 12 12.2 2.7 0.336
Adhesive (adhérant) Complexive astringency dynamic 9 18.4 4.0 0.801
Soft (doux) Complexive astringency complex 8 20.4 4.4 0.475
Mouthcoating (enrobant) Complexive astringency complex 10 16.3 3.5 0.928
Aggressive (agressif) Complexive astringency complex 2 32.7 7.1 0.071

a Sub-qualities grouping taken from Gawel et al. (2000). b p-value according to the Cochran’s Q test for descriptors in discrimination of samples. Values in bold indicate relevant terms based on Cochran’s Q test (p < 0.1) or rank of citation (1st to 8th most cited terms).

The frequency table shows the most cited terms and the most frequently used “group” to describe the selected grape extracts. The frequencies were also analysed in a Correspondence Analysis (CA, Figure 2) to find out if these terms differentiated the selected fractions. A combination of the most cited attributes and their ability to discriminate samples was taken into consideration for further sensory assessment of grape extracted fractions.

The three most frequently cited descriptors were “bitter” (13.3 %, 1st), “aggressive” (7.1 %, 2nd), and “salty” (6.6 %, 3rd) (Table 4). Interestingly, “bitter” was often reported, but did not allow the samples to be differentiated (p = 0.464). In contrast, “salty” and “aggressive” were relevantly used for the GF and SdTOT sample characterisation respectively, and showed better sample differentiation (p = 0.045 and p = 0.071, for “salty” and “aggressive” respectively). These descriptors, together with “acid” and “hot” (6.2 %, 4th) can be explained in part by the model wine solution, since tartaric acid and low pH are in accordance with acidic traits. Additionally, ethanol has been reported to be perceived as bitter and to induce a hot sensation (Vidal et al., 2004b). Likewise, phenolic compounds have been reported as bitter markers (Gonzalo-Diago et al., 2014; Hufnagel and Hofmann, 2008), and they therefore are common descriptors for grape extracts.

Terms for astringent sub-qualities accounted for 22.1 % of the total citations, the terms for “dusty” and “emery” being the most cited with 5.8 % (5th) and 5.3 % (6th) of the total citation frequencies. Furthermore, “chalky” and “silky” were frequently cited (4.4 %, 8th and 4.0%, 9th respectively; Table 4).

The CA biplot is shown in Figure 2. The two first dimensions account for 81.0 % of the explained variance among samples. Dimension 1 (Dim1) contributed 46.3 %, and it was mainly described by astringency sub-quality attributes: “silky” was negatively correlated (0.805) and contributed to sample discrimination (p = 0.040), whereas “emery” mainly contributed to the positive loading (+ 0.394, p > 0.1). In general, samples on the right side of the plot can be considered to have harsher astringency sub-qualities (“emery”), whereas samples plotted on the left side evoked softer sensations (“silky”). This is in accordance with the “aggressive” attribute, which characterised the right side of the axis (p = 0.071), in contrast with “oily” (p = 0.069) and “silky” (p = 0.040), which characterised the left side.

On the other hand, Dimension 2 (Dim2) accounted for 34.7 % of the explained variance, and was positively correlated with “watery” and “mouthcoating” attributes (+ 0.396 and + 0.345 respectively), and negatively correlated with “dry” and “grainy” attributes (- 0.995, p = 0.037 and - 0.504 respectively). Nevertheless, these attributes were not able to significantly discriminate samples even with increased significance criteria (p < 0.1). Moreover, the “mouthcoating”, “watery”, “grainy”, and “dry” attributes belong to different groups; i.e., complex (overall astringency), weight (mouthfeel), particulate (astringency sub-qualities), and drying (mouthfeel) respectively. Despite the variability of the group and the low discrimination ability (p > 0.1), the descriptors reported in the upper quadrant are linked by a general softer sensation with respect to those located in the lower quadrants of the plot. According to the plot description, Sk-TOT is related to harsher sensations, such as “grainy” and “dry”, whereas SdTOT corresponds to “aggressive” and “emery” attributes.

Regarding anthocyanin-based fractions, TA and GF are in the upper part of the plot, and are therefore represented by “watery” and “mouthcoating” sensations. The glucoside fraction (GF), however, is differentiated by softer astringency sub-qualities, such as “silky”, in contrast to total anthocyanins (TA), which are more associated with “emery”. Interestingly, SkTOT and GF were well differentiated from SdTOT, whereas TA was slightly different from GF and SkTOT, and more like SdTOT in the tasted concentrations.

In terms of both frequency and discrimination ability, the most cited term, “bitter”, was considered relevant enough to be investigated. Astringent sub-qualities such as surface-smoothness and particulate were also found to be both highly cited and able to discriminate the samples. Therefore, they were chosen to train assessors in using the scales. For example, particulate was described as “feelings of particulate matter brushing against the surface of the mouth through the movement of the solution”, thus as “in-mouth” astringency; whereas surface smoothness was defined as “texture felt on mouth surfaces when the different surfaces came in contact with each other”, therefore after expectoration. Reference standards and scales were defined, slightly modifying the scales proposed by Pickering and Demiglio (2008), for the evaluation of oral sensations elicited by white wine. Overall astringency was chosen for the scales to summarise the intensity of both in-mouth and post expectoration astringency.

Moreover, given the high citation frequency of the term “acid” and “hot” (6.2 %, 4th) and also “tingly” and “prickly” (together 7.5 %), the panel discussed and attempted to scale these sensations and to verbally explain them. In this case, the Pickering and Demiglio (2008) definition of “tingly” was once again used, and the scale was divided into low, moderate and high irritation.

Figure 2. Correspondence Analysis (CA) of descriptors for the selected grape extracts performed on the citation frequencies.

GF = glucoside anthocyanin fraction; TA = total anthocyanin fraction; SdTOT = seed polyphenol extract; SkTOT = skin polyphenol extract. Ellipses represent confidence ellipses (p = 0.05).

Figure 2. Correspondence Analysis (CA) of descriptors for the selected grape extracts performed on the citation frequencies.

GF = glucoside anthocyanin fraction; TA = total anthocyanin fraction; SdTOT = seed polyphenol extract; SkTOT = skin polyphenol extract. Ellipses represent confidence ellipses (p = 0.05).

3. DA results

Panel performance was evaluated using a PCA approach (data not shown) to understand the consensus in the attributes investigated and ANOVA for the effects of assessors, extracts and replicates and their interactions in order to assess panel repeatability. Regarding consensus, assessor projections were grouped in the loading plot for overall astringency, particulate, and surface smoothness, showing that the panel had agreed on the interpretation of these terms. For irritation, however, the assessors were spread over the loading plot, which suggests that the assessors did not interpret this attribute in the same way. In addition, for the terms involved, the ANOVA test for assessors resulted in significant differences (p < 0.05), underlining the differences in the evaluation of these attributes on the scale. Hence, irritation was no longer considered in subsequent analyses, confirming that the use of standard references should be preferred to verbal description of terms. ANOVA also showed that there was a significant assessors* -sample interaction (p < 0.05) for the attribute bitterness. Examination of the PCA projection indicated that two assessors rated bitterness differently from the rest of the panel. Accordingly, none of the scores from these two assessors were taken into consideration for further analysis. Therefore, the final DA panel consisted of ten assessors.

3.1. Grape extract sensory characterisation

The result of the ANOVA test for the extract effect (Figure 3) shows that overall astringency, particulate, and surface smoothness were significantly different among extracts ( p = 0.003, p = 0.001, and p = 0.002, respectively), but not bitterness (p > 0.05). Likewise, Brossaud et al. (2001) did not find significant differences between total anthocyanin extract and tannins from seeds and skins added to white wine, whereas differences were reported in citric acid solutions, in which seed extracts were significantly higher than skin and anthocyanin extracts. Our different results can be imputed to differences in polyphenol extracts. In fact, different content and composition of polyphenols in skin and seed extracts can vary depending on the grape features and the purification steps applied. Alcohol can also enhance bitterness (Mattes and DiMeglio, 2001), which may conceal individual differences between fractions.

Figure 3. Descriptive analysis (DA) of grape extracts.

SdTOT = seed polyphenol extract; SkTOT = skin polyphenol extract; TA = total anthocyanin extract; GF = glucoside anthocyanin fraction; AF = acetylated anthocyanin fraction; CF = p-coumaroylated anthocyanin fraction. Data are expressed as means of assessor ratings and error bars are calculated as s/(n)1/2 , where s is standard deviation and n is the number of assessors. p-values are reported according to ANOVA, and different Latin letters indicate significant differences according to the HSD Tukey test (p < 0.05).

Figure 3. Descriptive analysis (DA) of grape extracts.

SdTOT = seed polyphenol extract; SkTOT = skin polyphenol extract; TA = total anthocyanin extract; GF = glucoside anthocyanin fraction; AF = acetylated anthocyanin fraction; CF = p-coumaroylated anthocyanin fraction. Data are expressed as means of assessor ratings and error bars are calculated as s/(n)1/2 , where s is standard deviation and n is the number of assessors. p-values are reported according to ANOVA, and different Latin letters indicate significant differences according to the HSD Tukey test (p < 0.05).

In contrast, overall astringency differed among the fractions: SkTOT was the highest in astringency, but not significantly different from SdTOT and TA. Anthocyanin fractions (GF, AF, and CF) were rated lower in overall astringency, significantly less than SkTOT, but not different from SdTOT and TA. In previous research, anthocyanin astringency ratings have been found to be lower than seed or skin extracts, and differences have been found depending on the media in which they were tasted; i.e., model wine at different ethanol level, citric acid solution, and white wine (Brossaud et al., 2001; Vidal et al., 2004b). Regarding GF, anthocyanin glucoside has been reported to have moderate astringency (Ferrer-Gallego et al., 2015).

Interestingly, the astringency sub-qualities, particulate and surface smoothness, differed among samples, and they agreed with the overall perceived astringency results. In fact, SkTOT was also perceived as the highest on the scale of sub-qualities, corresponding to anchored reference, “fine emery”, for surface smoothness, and close to “grainy” for particulate attribute. This could agree with higher mean polymerisation (mDP) of skin tannins, which increases both the overall astringency and the puckery sensation in the mouth (Ma et al., 2014). TA, GF, and CF fractions did not differ from polyphenol extracts in particulate attribute, although they were rated lower. Total anthocyanins added to white juice have previously been reported to result in a wine with an increased “fine grain” attribute (Oberholster et al., 2009). Our results seem to agree with the CATA test, since TA rating corresponds to “dusty” and “sandy” on the scale, referring to the moderate particulate sensations in the mouth. The acetylated fraction (AF) was rated the lowest and was significantly different from SkTOT extract, evoking a “chalky” sensation previously reported in literature in an anthocyanin sensory analysis (Vidal et al., 2004a; Vidal et al., 2004b). Surface smoothness of AF and CF was significantly lower than that of skin polyphenol extracts (p = 0.002). According to the scale, AF and CF were closer to a “silky” sensation and TA and GF correspond to a “velvety” one, whereas SkTOT corresponded to “fine emery”. Interestingly, a previous study (Ferrer-Gallego et al., 2015) reported an increase in the descriptor “velvety” in wine with added anthocyanin glucoside (400 mg/L), which is in line with the results of present study.

3.2. Mixed fractions sensory analysis

Anthocyanin fractions tasted alone were perceived as contributing to bitterness, overall astringency and its sub-qualities, surface smoothness and particulate to different extents, depending on their compositions and their concentrations. Although these contributions had lower ratings with respect to other phenol fractions, it is interesting to determine whether the sensory perceptions of polyphenol extracts are influenced by the addition of total anthocyanin extracts (TA) and their derived fractions GF and CF (Table 5). In previous research, total anthocyanin extracts were found to increase overall astringency when added to seed or skin extracts (Brossaud et al., 2001; Vidal et al., 2004b), although this increase depends on the concentration of proanthocyanidins (Vidal et al., 2004b). Under our experimental conditions, when anthocyanin fractions were added to the SkTOT extract, no significant differences (p > 0.05) were found for any of the investigated attributes, except for surface smoothness, which was rated significantly lower (p = 0.043) in the sample with added GF. Conversely, a significant increase was found when GF was added to SdTOT for the particulate and surface smoothness sub-qualities (p = 0.010 and p = 0.030 respectively), as well as for overall astringency (p = 0.009). Depending on the polyphenol fraction, surface smoothness descriptors showed an inverse trend when anthocyanins were added, decreasing when glucoside anthocyanins were added to skin polyphenols (p < 0.05) and increasing when they were added to seed polyphenols (p < 0.05). Recently, Soares et al. (2019) showed that the mixture of epicatechin and malvidin-3-glucoside increased the interaction with salivary proline-rich proteins when compared to epicatechin alone, which may result in an increased perceived astringency, and thus justify the results obtained here for SdTOT. Nevertheless, this increase could be hidden in the skin polyphenol extract, because of the lower amount of monomeric flavanols usually reported in skin with respect to seeds. The condensed tannins of skins are conversely known to be highly polymerised, and increased polymerisation is positively correlated with perceived astringency (Ma et al., 2014). In fact, condensed tannins are the first drivers of in-mouth related sensations, particularly mouthfeel (Ferrero-del-Teso et al., 2020). However, these results confirm the involvement - even if to a different extent - of other relevant groups of polyphenols (such as anthocyanins) in the final oro-sensory perception of wine, as has already been suggested by Ferrero-del-Teso et al. (2020).

Table 5. Descriptive Analysis (DA) of mixed extracts.
titre du tableau
Mixed extracts Bitterness Particulate Surface Smoothness Overall Astringency
SkTOT 3.6 ± 0.5 4.4 ± 0.5 4.6 ± 0.4 a 4.1 ± 0.5
SkTOT + TA 3.7 ± 0.5 4.3 ± 0.6 4.1 ± 0.5 ab 3.9 ± 0.5
SkTOT + GF 5.0 ± 0.5 5.3 ± 0.4 3.8 ± 0.3 b 5.4 ± 0.4
SkTOT + CF 4.3 ± 0.4 4.6 ± 0.5 4.0 ± 0.6 ab 4.0 ± 0.7
Sign ns ns * ns
p value 0.201 0.441 0.043 0.077
SdTOT 4.2 ± 0.5 3.5 ± 0.5 b 3.6 ± 0.6 b 3.0 ± 0.4 b
SdTOT + TA 5.2 ± 0.4 4.0 ± 0.5 ab 4.0 ± 0.4 ab 3.6 ± 0.3 ab
SdTOT + GF 3.8 ± 0.4 5.3 ± 0.5 a 5.1 ± 0.4 a 4.9 ± 0.5 a
SdTOT + CF 4.5 ± 0.6 3.3 ± 0.4 b 4.1 ± 0.4 ab 3.4 ± 0.4 b
Sign ns * * **
p value 0.177 0.010 0.030 0.009

SdTOT = seed polyphenol extract; SkTOT = skin polyphenol extract; TA = total anthocyanin fraction; GF = glucoside anthocyanin fraction; CF = p-coumaroylated anthocyanin fraction. Data are expressed as means of assessors rating and error bars are calculated as s/(n)1/2 , where s is standard deviation and n is the number of assessors. Sign: *, **, and ns indicate significance at p < 0.05, 0.01, and not significant respectively, for the differences among samples according to ANOVA. Different Latin letters indicate significant differences according to HSD Tukey test (p < 0.05).

 Conclusions

The role of anthocyanins in the in-mouth sensory properties of wine has been evaluated in previous studies with contradictory results. The aim of the present study was to adapt common methodology for sensory analysis to the investigation of the contribution of anthocyanins alone and in more complex solutions, such as mixed polyphenol extracts. The results showed that anthocyanins extracted from grape skins are involved in the “in-mouth” perception of model wine solutions, although their contribution is less relevant than other polyphenolic groups, such as condensed tannins. At the wine level, glucoside anthocyanins were related to sensations of “velvety” sub-qualities, whereas when mixed with the other acylation groups (total anthocyanin extract) they showed higher overall astringency and harsher sub-qualities than when alone. In particular, glucoside anthocyanins led to a decrease in overall astringency in the presence of skin polyphenols, whereas they seemed to enhance the overall astringency and related sub-qualities of seed polyphenols. This behaviour could be relevant when explaining the final in-mouth perceptions of wine, and thus further studies would be required on the different interactions of fractions with a suppression or enhancing effects of sensory attributes.

Acknowledgements

M.A.P. would like to thank Università Italo-Francese (Torino, Italy) for the Vinci Grant 2017 supporting co-jointed PhD.

Brossaud F. , Cheynier V. , Noble A. C. , 2001. Bitterness and astringency of grape and wine polyphenols. Australian Journal of grape and wine research, 7(1), 33-39. doi.org/10.1111/j.1755-0238.2001.tb00191.x. Campo E. , Do B.V. , Ferreira V. , Valentin D. , 2008. Aroma properties of young Spanish monovarietal white wines: A study using sorting task, list of terms and frequency of citation. Australian Journal of Grape and Wine Research, 14 (2008), pp. 104-115. doi.org/10.1111/j.1755-0238.2008.00010.x. Chira K. , Jourdes M. , Teissedre P. L. , 2012. Cabernet-Sauvignon red wine astringency quality control by tannin characterization and polymerization during storage. European Food Research and Technology, 234(2), 253-261. doi.org/10.1007/s00217-011-1627-1. Ferrer-Gallego R. , Soares S. , Mateus N. , Rivas-Gonzalo J. , Escribano-Bailón M. T. , Freitas V. D. , 2015. New anthocyanin–human salivary protein complexes. Langmuir, 31(30), 8392-8401. doi.org/10.1021/acs.langmuir.5b01122. Ferrero-del-Teso S. , Suárez A. , Jeffery D. W. , Ferreira V. , Fernández-Zurbano P. , Sáenz-Navajas M. P. , 2020. Sensory variability associated with anthocyanic and tannic fractions isolated from red wines. Food Research International, 109340. doi.org/10.1016/j.foodres.2020.109340. Fontoin H. , Saucier C. , Teissedre P. L. , Glories Y. , 2008. Effect of pH, ethanol and acidity on astringency and bitterness of grape seed tannin oligomers in model wine solution. Food Quality and Preference, 19(3), 286-291. doi.org/10.1016/j.foodqual.2007.08.004. Fukui M. , Yokotsuka K. , Ishii R. , O'Mahony M. , Rousseau B. , 2002. Investigation of potential taste reduction of catechin and grape seed dimeric phenols in water by wine proteins. LWT-Food Science and Technology, 35(4), 355-361. doi.org/10.1006/fstl.2001.0887. Gawel R. , 1997. The use of language by trained and untrained experienced wine tasters. Journal of Sensory Studies, 12(4), 267-284. doi.org/10.1111/j.1745-459X.1997.tb00067.x. Gawel R. , Oberholster A. , Francis I. L. , 2000. A ‘Mouth-feel Wheel': terminology for communicating the mouth-feel characteristics of red wine. Australian Journal of Grape and Wine Research, 6(3), 203-207. doi.org/10.1111/j.1755-0238.2000.tb00180.x. Gibbins H. L. , Carpenter G. H. , 2013. Alternative mechanisms of astringency - what is the role of saliva? Journal of Texture Studies, 44(5), 364-375. doi.org/10.1111/jtxs.12022. Glabasnia A. , Hofmann T. , 2006. Sensory-directed identification of taste-active ellagitannins in American (Quercus alba L.) and European oak wood (Quercus robur L.) and quantitative analysis in bourbon whiskey and oak-matured red wines. Journal of Agricultural and Food Chemistry, 54(9), 3380-3390. doi.org/10.1021/jf052617b. González-Centeno M. R. , Jourdes M. , Femenia A. , Simal S. , Rosselló C. , Teissedre P. L. , 2012. Proanthocyanidin composition and antioxidant potential of the stem winemaking byproducts from 10 different grape varieties (Vitis vinifera L.). Journal of Agricultural and Food Chemistry, 60(48), 11850-11858. doi.org/10.1021/jf303047k. Gonzalo-Diago A. , Dizy M. , Fernández-Zurbano P. , 2014. Contribution of low molecular weight phenols to bitter taste and mouthfeel properties in red wines. Food Chemistry, 154, 187-198. doi.org/10.1016/j.foodchem.2013.12.096. Harrison R. , 2018. Practical interventions that influence the sensory attributes of red wines related to the phenolic composition of grapes: a review. International Journal of Food Science & Technology, 53(1), 3-18. doi.org/10.1111/ijfs.13480. Hothorn T. , Bretz F. , Westfall P. , 2008. Simultaneous Inference in General Parametric Models. Biometrical Journal, 50(3), 346–363. doi.org/10.1002/bimj.200810425. Hufnagel J. C. , Hofmann T. , 2008. Orosensory-directed identification of astringent mouthfeel and bitter-tasting compounds in red wine. Journal of Agricultural and Food Chemistry, 56(4), 1376-1386. doi.org/10.1021/jf073031n. Laguna L. , Bartolomé B. , Moreno-Arribas M. V. , 2017. Mouthfeel perception of wine: Oral physiology, components and instrumental characterization. Trends in Food Science & Technology, 59, 49-59. doi.org/10.1016/j.tifs.2016.10.011. Lê S. , Josse J. , Husson F. , 2008. FactoMineR: An R package for multivariate analysis. Journal of Statistical Software. 25(1). pp. 1-18.. Lorrain B. , Chira K. , Teissedre P. L. , 2011. Phenolic composition of Merlot and Cabernet-Sauvignon grapes from Bordeaux vineyard for the 2009-vintage: comparison to 2006, 2007 and 2008 vintages. Food Chemistry, 126(4), 1991-1999. doi.org/10.1016/j.foodchem.2010.12.062. Ma W. , Guo A. , Zhang Y. , Wang H. , Liu Y. , Li H. , 2014. A review on astringency and bitterness perception of tannins in wine. Trends in Food Science & Technology, 40(1), 6-19. doi.org/10.1016/j.tifs.2014.08.001. Ma W. , Waffo-Teguo P. , Jourdes M. , Li H. , Teissedre P. L. , 2016. Chemical affinity between tannin size and salivary protein binding abilities: Implications for wine astringency. PloS one, 11(8). doi.org/10.1371/journal.pone.0161095. Mattes R. D. , DiMeglio D. , 2001. Ethanol perception and ingestion. Physiology & Behavior, 72(1-2), 217-229. doi.org/10.1016/S0031-9384(00)00397-8. Oberholster A. , Francis I. L. , Iland P. G. , Waters E. J. , 2009. Mouthfeel of white wines made with and without pomace contact and added anthocyanins. Australian Journal of Grape and Wine Research, 15(1), 59-69. doi.org/10.1111/j.1755-0238.2008.00038.x. Paissoni M. A. , Waffo-Teguo P. , Ma W. , Jourdes M. , Rolle L. , Teissedre P. L. , 2018. Chemical and sensorial investigation of in-mouth sensory properties of grape anthocyanins. Scientific Reports, 8(1), 1-13. doi.org/10.1038/s41598-018-35355-x. PanelCheck software , 2012. V 1.4.2, Nofima Mat, Tromsø, Norway. http://www.panelcheck.com. Parpinello G. P. , Versari A. , Chinnici F. , Galassi S. , 2009. Relationship among sensory descriptors, consumer preference and color parameters of Italian Novello red wines. Food Research International, 42(10), 1389-1395. doi.org/10.1016/j.foodres.2009.07.005. Parr W. V. , Heatherbell D. , White K. G. , 2002. Demystifying wine expertise: Olfactory threshold, perceptual skill and semantic memory in expert and novice wine judges. Chemical senses, 27(8), 747-755. doi.org/10.1093/chemse/27.8.747. Peynaud E. , 1987. The taste of wine: The art and science of wine appreciation. Macdonald Orbis, London. ISBN 13: 9780471113768. Pickering G. J. , Demiglio P. , 2008. The white wine mouthfeel wheel: A lexicon for describing the oral sensations elicited by white wine. Journal of Wine Research, 19(1), 51-67. doi.org/10.1080/09571260802164038. Pinheiro J. , Bates D. , DebRoy S. , Sarkar D. , R Core Team , 2020. nlme: Linear and Nonlinear Mixed Effects Models. https://CRAN.R-project.org/package=nlme.. Piombino P. , Pittari E. , Gambuti A. , Curioni A. , Giacosa S. , Mattivi F. , Parpinello G.P. , Rolle L. , Ugliano M. , Moio L. , 2020. Preliminary sensory characterization of the diverse astringency of mono-varietal Italian red wines and correlation of sub-qualities with chemical parameters. Australian Journal of. Grape and Wine Research, 26(3), 233-246. doi.org/10.1111/ajgw.12431. R Core Team , 2018. R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. https://www.R-project.org/ (2018).. RStudio Team , 2018. RStudio: Integrated Development for R. RStudio, PBC, Boston, (MA), USA. http://www.rstudio.com/.. Sáenz-Navajas M. P. , Avizcuri J. M. , Echávarri J. F. , Ferreira V. , Fernández-Zurbano P. , Valentin D. , 2016. Understanding quality judgements of red wines by experts: Effect of evaluation condition. Food Quality and Preference, 48, 216-227. doi.org/10.1016/j.foodqual.2015.10.001. Sáenz-Navajas M. P. , Avizcuri J. M. , Ferrero-del-Teso S. , Valentin D. , Ferreira V. , Fernández-Zurbano P. , 2017. Chemo-sensory characterization of fractions driving different mouthfeel properties in red wines. Food Research International, 94, 54-64. doi.org/10.1016/j.foodres.2017.02.002. Singleton V.L. , Noble A.C. , 1976. Wine flavor and phenolic substances. In: Phenolic, sulfur and nitrogen compounds in food flavours. Eds G. Charalambous and I. Katz, American Chemical Society Symposium Series, 26, pp. 47–70. ISBN 13: 9780841203303. Soares S. , Ferrer-Galego R. , Brandão E. , Silva M. , Mateus N. , de Freitas V. , 2016. Contribution of human oral cells to astringency by binding salivary protein/tannin complexes. Journal of Agricultural and Food Chemistry, 64(41), 7823-7828.. Soares S. , Kohl S. , Thalmann S. , Mateus N. , Meyerhof W. , De Freitas V. , 2013. Different phenolic compounds activate distinct human bitter taste receptors. Journal of Agricultural and Food chemistry, 61(7), 1525-1533. doi.org/10.1021/acs.jafc.6b02659. Soares S. , Silva M. S. , García-Estévez I. , Brandão E. , Fonseca F. , Ferreira-da-Silva F. , Escribano-Bailón M.T. , Mateus N. , de Freitas V. , 2019. Effect of malvidin-3-glucoside and epicatechin interaction on their ability to interact with salivary proline-rich proteins. Food Chemistry, 276, 33-42. doi.org/10.1016/j.foodchem.2018.09.167. Soares S. , Soares S. , Brandão E. , Guerreiro C. , Mateus N. , de Freitas V. , 2020. Oral interactions between a green tea flavanol extract and red wine anthocyanin extract using a new cell-based model: insights on the effect of different oral epithelia. Scientific Reports, 10(1), 1-16. doi.org/10.1038/s41598-020-69531-9. Tomic O. , Luciano G. , Nilsen A. , Hyldig G. , Lorensen K. , Næs T. , 2010. Analysing sensory panel performance in a proficiency test using the PanelCheck software. European Food Research and Technology, 230(3), 497. DOI 10.1007/s00217-009-1185-y. Varela P. , Ares G. , 2012. Sensory profiling, the blurred line between sensory and consumer science. A review of novel methods for product characterization. Food Research International, 48(2), 893-908. doi.org/10.1016/j.foodres.2012.06.037. Vidal S. , Francis L. , Noble A. , Kwiatkowski M. , Cheynier V. , Waters E. , 2004. a. Taste and mouth-feel properties of different types of tannin-like polyphenolic compounds and anthocyanins in wine. Analytica Chimica Acta, 513 (1), 57-65. doi.org/10.1016/j.aca.2003.10.017. Vidal S. , Courcoux P. , Francis L. , Kwiatkowski M. , Gawel R. , Williams P. , Waters E. , Cheynier V. , 2004. b. Use of an experimental design approach for evaluation of key wine components on mouth-feel perception. Food Quality and Preference, 15(3), 209-217. doi.org/10.1016/S0950-3293(03)00059-4. Vidal S. , Francis L. , Williams P. , Kwiatkowski M. , Gawel R. , Cheynier V. , Waters E. , 2004. c. The mouth-feel properties of polysaccharides and anthocyanins in a wine like medium. Food Chemistry, 85(4), 519-525. doi.org/10.1016/S0308-8146(03)00084-0.