Food Chemistry 154 (2014) 187–198

Contents lists available at ScienceDirect

Food Chemistry

journal homepage: www.elsevier .com/locate / foodchem

Contribution of low molecular weight phenols to bitter tasteand mouthfeel properties in red wines

0308-8146/$ - see front matter � 2014 Elsevier Ltd. All rights reserved.http://dx.doi.org/10.1016/j.foodchem.2013.12.096

⇑ Corresponding author at: Instituto de Ciencias de la Vid y el Vino (UR-CSIC-GR),Madre de Dios 51, E-26006 Logroño, La Rioja, Spain. Tel.: +34 941299622; fax: +34941299621.

E-mail address: puri.fernandez@unirioja.es (P. Fernández-Zurbano).

Ana Gonzalo-Diago a,b, Marta Dizy a,c, Purificación Fernández-Zurbano a,b,⇑a Instituto de Ciencias de la Vid y el Vino (UR-CSIC-GR), Madre de Dios 51, E-26006 Logroño, La Rioja, Spainb Department of Chemistry, University of La Rioja, Madre de Dios 51, E-26006 Logroño, La Rioja, Spainc Department of Agriculture and Food, University of La Rioja, Madre de Dios 51, E-26006 Logroño, La Rioja, Spain

a r t i c l e i n f o

Article history:Received 29 August 2013Received in revised form 23 December 2013Accepted 29 December 2013Available online 7 January 2014

Keywords:WinePolyphenolsSensory analysisBitterAstringency

a b s t r a c t

The aim of this study was to explore the relationship between low molecular weight compounds presentin wines and their sensory contribution. Six young red wines were fractionated by gel permeation chro-matography and subsequently each fraction obtained was separated from sugars and acids by solid phaseextraction. Wines and both fractions were in-mouth evaluated by a trained sensory panel and UPLC–MSanalyses were performed. The lack of ethanol and proanthocyanidins greatly increased the acidityperceived. The elimination of organic acids enabled the description of the samples, which were evaluatedas bitter, persistent and slightly astringent. Coutaric acid and quercetin-3-O-rutinoside appear to berelevant astringent compounds in the absence of proanthocyanidins. Bitter taste was highly correlatedwith the in-mouth persistence. A significant predictive model for bitter taste was built by means of PLSR.Further research must be carried out to validate the sensory contribution of the compounds involved inbitterness and astringency and to verify the sensory interactions observed.

� 2014 Elsevier Ltd. All rights reserved.

1. Introduction

The comprehensive study of non-volatile compounds in redwine is of great interest due to the sensory properties of thesecompounds, such as sweetness, acidity, bitterness and differentoral mouthfeel perceptions such as velvety, puckering and dryingastringency, among others. The quality perception of a wine is dri-ven primarily by the absence of defective aroma and secondarily tothe presence of non-volatile components and more precisely tophenolic composition that is able to modulate quality perception(Sáenz-Navajas, Tao, Dizy, Ferreira, & Fernández-Zurbano, 2010).

The contribution of non-volatile molecules to wine sensoryproperties has been widely published (Arnold, Noble, & Singleton,1980; Chira, Pacella, Jourdes, & Teissedre, 2011; Gawel, Francis, &Waters, 2007; Landon, Weller, Harbertson, & Ross, 2008; Peleg,Gacon, Schlich, & Noble, 1999; Preys et al., 2006; Sáenz-Navajas,Avizcuri, Ferreira, & Fernández-Zurbano, 2012; Sáenz-Navajas,Campo et al., 2012; Vidal, Courcoux et al., 2004) and most publica-tions have studied in detail the compounds that contribute most toastringency perception (Chira et al., 2011; Gawel et al., 2007;Landon et al., 2008; Sáenz-Navajas, Avizcuri et al., 2012). Withthis purpose, Hufnagel and Hofmann (2008a) carried out

reconstruction studies from the nonvolatile composition of a redwine, demonstrating that puckering astringency (using tannic acidas reference standard) is mainly caused by a polymeric fractionexhibiting molecular masses above 5 kDa, this oral sensation beingamplified by acids such as caftaric acid, gallic acid and furan-2-carboxilic acid. Other study performed with the same goal(Sáenz-Navajas, Avizcuri et al., 2012) developed two models forpredicting perceived astringency (using, in this case, potassiumand aluminium sulphate as the reference standard for astringency).In both models, the concentration of proanthocyanidins, thepresence of organic acids and also ethanol content once again ac-counted for perceived astringency. Monomeric phenols have beenrepeatedly described as astringent and bitter (Arnold et al., 1980;Hufnagel & Hofmann, 2008a; Peleg et al., 1999), although recentstudies have shown that monomeric phenols are not present inconcentrations above their sensory threshold, suggesting thatthese compounds might not play an important role in the sensoryperception of red wines (Hufnagel & Hofmann, 2008a; Sáenz-Navajas,Avizcuri et al., 2012).

In contrast, few authors have focused on the study of bittertaste in red wines, with controversy surrounding the results ob-tained for the compounds eliciting bitter taste (Arnold et al.,1980; Hufnagel & Hofmann, 2008a; Kallithraka, Bakker, & Clifford,1997; Peleg et al., 1999; Robichaud & Noble, 1990). Furthermore,some authors, despite training assessors specifically in bitter term,have reported differences in its interpretation (Sáenz-Navajas,

188 A. Gonzalo-Diago et al. / Food Chemistry 154 (2014) 187–198

Avizcuri et al., 2012), while other authors have reported a very sig-nificant sample effect for each attribute studied except for bitter-ness (Vidal et al., 2003). Several authors (Peleg et al., 1999;Robichaud & Noble, 1990) have studied the bitterness of polyphe-nol compounds, such as polymeric fractions of tannic acid and tan-nins, as well as flavan-3-ol monomers, dimers, and trimers,demonstrating that larger molecules tend to be less bitter andmore astringent. Peleg et al. (1999) found that (�)-epicatechinwas more bitter than the stereoisomer (+)-catechin and that bothwere more bitter than the procyanidin trimers, catechin-(4–8)-cat-echin-(4–8)-catechin and catechin-(4–8)-catechin-(4–8)-epicate-chin, in contrast to Hufnagel and Hofmann (2008a) found thatprocyanidins dimers and a procyanidin trimer were more bitterthan (�)-epicatechin and catechin. One study focused on whitewines with and without pomace contact and with the addition ofanthocyanins (Oberholster, Francis, Iland, & Waters, 2009), estab-lishing that the score for bitterness attribute was correlated withthe concentration of most phenolic compounds, but especially withproanthocyanidins and polymeric phenols, which also coincidedwith previous reports (Arnold et al., 1980). In contrast, Hufnageland Hofmann (2008a) considered polymeric phenols (>5 kDa) asnon-bitter compounds. On the other hand, studies carried out onwhite wines and model solutions have demonstrated that catechinelicits both bitterness and astringency (Arnold et al., 1980;Robichaud & Noble, 1990). Sáenz-Navajas, Ferreira, Dizy, andFernández-Zurbano (2010) studying red wine fractions reportedthat bitterness might be explained by the presence of monomerssuch as catechin and epigallocatechin, phenolic acids such ascoutaric and caftaric acid and flavonols such as myricetin. Withthe same goal in mind, Hufnagel and Hofmann (2008a) by meansof taste reconstruction and omission experiments described aspotential bitter compounds two flavan-3-ol monomers and fourdimers, seven phenolic acids and eight amino acids. Althoughthe concentration of these compounds was ten times belowtheir threshold concentrations, Hufnagel and Hofmann (2008a)concluded that sub-threshold concentrations of phenolic acid ethylesters and flavanols contribute to red wine bitterness. In spite ofthese inconsistencies, not enough is known about the bitter tasteof polyphenol compounds that do not belong to tannin classes,such as anthocyanins.

Another important aspect to take into account is the fact thathuman ability to taste bitterness is genetically dependent andapproximately 30% of population is taste blind to the bitternessof bitter synthetic compounds such as phenylthiocarbamide(PTC) and 6-n-propylthiouracil (PROP) (Tepper et al., 2009). Thisgenetic ability is probably linked to the presence of bitter-tastingcompounds in sub-threshold concentrations, and both issues maybe the key to understanding why bitter taste is still far from beingunderstood.

It has also been reported that not only chemical compositionbut also molecular interactions among wine components play adeterminant role in the chemical stability of wine, also affectingits sensory properties (Sáenz-Navajas, Campo, Fernández-Zurbano,Valentin, & Ferreira, 2010). It has been extensively reported thatethanol enhances perceived bitterness (Fischer & Noble, 1994;Noble, 1990; Oberholster et al., 2009; Vidal, Courcoux et al.,2004), masks it (Vidal, Francis, Noble et al., 2004) and can suppressthe astringency of phenols (Noble, 1990; Vidal, Courcoux et al.,2004). An increase of 3% v/v of ethanol increases bitterness more(by nearly 50%) than the addition of 1400 mg L�1 of catechin tothe same wine (which increased bitterness by only 28%) (Fischer& Noble, 1994). Increased acidity (and perceived sourness)increased the intensity of astringency (Kallithraka et al., 1997).

The possibility of identifying relationships between composi-tion and sensory description will provide more information to-wards a better understanding of how interactions between

chemical components may affect flavour perception. The generalaim of this study was to advance in the knowledge of the effectof non-volatile low molecular weight phenolic compounds on in-mouth taste and feeling perceptions, especially bitterness andastringency. The specific aims of this study were: (1) to study cor-relations between in-mouth sensory properties and low molecularweight non-volatile compounds; and (2) to explore the role of lowmolecular weight non-volatile compounds in the bitter taste of redwines. To achieve both goals, six Spanish young red wines showingdifferent total polyphenol index values and subsequently the frac-tions obtained from them were in-mouth sensory described by atrained panel selected for its ability to taste bitter. The non-volatilecompounds were identified and quantified by UPLC–DAD–MS. Thepotential contribution of non-volatile molecules to bitter taste wasstudied using statistical tools.

2. Materials and methods

2.1. Chemicals and reagents

All chemicals used were of analytical reagent grade. All chro-matographic solvents were of HPLC grade. Ultrapure water was ob-tained from a Milli-Q purification system (Millipore, Molsheim,France). Spring water was purchased from Solán de Cabras (Cuen-ca, Spain). Methanol, formic acid, ethanol, acetonitrile, sulphuricand hydrochloric acid were purchased from Scharlab (Barcelona,Spain). Quinine sulphate dihydrate (98%) was obtained from AlfaAesar (Karlsruhe, Germany). Potassium and aluminium sulphateand tannic acid were purchased from Panreac (Barcelona, Spain).L-Tartaric acid, L-malic acid, L-lactic acid, succinic acid, citric acid,trans-aconitic acid, cis-aconitic acid, syringic acid, 6-propyl-2-thio-uracil, catechin, epicatechin, myricetin, kaempferol, vanillin,protocatechuic acid ethyl ester, protocatechuic acid, gallic acid,caffeic acid, resveratrol and quercetin were purchased fromSigma–Aldrich (St. Louis, MO, USA). Oenin-chloride, caffeic acidethyl ester, quercetin-3-O-glucoside, quercetin-3-O-rutinoside,quercetin-3-O-galactoside, quercetin-3-O-glucuronide, kaempferol-3-O-glucoside, isorhamnetin-3-O-glucoside, myricetin, kaempferol,isorhamnetin, epicatechin gallate, epigallocatechin, procyanidins A2,B1 and B2, ferulic and p-coumaric acids were provided by Extrasyn-these (Genay, France). Vanillic acid was supplied by Fluka (Buchs,Switzerland).

2.2. Wines

Six young red commercial wines were selected out of 35 winesfrom different Spanish Denominations of Origin and regions. Theselection was made on the basis of differences in their total poly-phenol index (TPI) values and also on the basis of different valuesgiven to bitterness and astringency attributes.

2.3. Sample preparation

2.3.1. Elimination of wine volatile compoundsWines were de-alcoholized and de-aromatized according to

Sáenz-Navajas, Campo et al. (2010) in order to obtain an odourlesstastant fraction from each wine. The non-volatile extract obtainedfrom 50 ml of wine was then re-dissolved in 2 ml of ethanol/water(13:87, v/v) in order to obtain the low molecular weight fractionsas follows.

2.3.2. Isolation of low molecular weight compoundsTSK Toyopearl gel HW-50F (Tosohaas, Montgomery-ville, PA,

USA) was suspended in miliQ water and, after swelling, it waspacked in a Millipore (Bedford, MA, USA) Vantage L column

A. Gonzalo-Diago et al. / Food Chemistry 154 (2014) 187–198 189

(120 mm � 12 mm i.d.) at atmospheric pressure. The system usedwas an Agilent modular 1100 liquid chromatograph (Waldbronn,Germany) equipped with a peristaltic pump (Agilent 61311A), aRheodyne injector (2 ml loop), a diode array detector (Agilent,G1315D) and Agilent Chemstation software. Two milliliters of thenon-volatile extract were directly applied to the column at a flowrate of 2 ml min�1. The fractionation was performed according tothe method described by Gonzalo-Diago, Dizy, and Fernández-Zurbano (2013). The low molecular weight compounds (sugars,anthocyanins, phenolic acids, organic acids, flavonols and flava-nols) were eluted with 240 ml of ethanol/water/formic acid(55:45:1 v/v/v) and manually collected in round flasks. Beforestarting another sample fractionation, the column was cleanedwith 40 ml of acetone/water (60:40, v/v) to wash the highmolecular weight compounds. The ethanol was evaporated undervacuum and the low molecular weight fraction (LMW-F) wasfreeze-dried and then re-dissolved in 32 ml of spring water forsensory analysis. For chemical analysis, the wine fraction wasreconstituted to its original concentration. In order to have enoughvolume to give to the panellists, 350 ml of wine was used per sam-ple; this fractionation was repeated seven times for each wine(50 ml per fractionation). LMW fractions were stored at 4 �C untilchemical and sensory analyses.

2.3.3. Solid-phase extractionA Bond Elut LRC C18 cartridges (500 mg, 10 ml) from Varian

(Darmsdtadt, Germany) were used. SPE was carried out to removeacids from the LMW fractions. For this purpose, one LMW fractionof each wine was used to obtain a LMW fraction without acids(LMWW/OA). SPE was performed according to the method describedby de Villiers, Lynen, Crouch, and Sandra (2004), with certain mod-ifications. Methanol was used instead of ethyl acetate to recoverthe phenolic fraction and 7 ml of sample were loaded in the car-tridge instead of one. This breakthrough volume was selected asthe best in a range from 5 to 15 ml. The cartridges were precondi-tioned with 5 ml of methanol followed by 5 ml of water acidified topH 2.5 with 1 M HCl. For this analysis, the LMW fraction was re-dissolved in 50 ml of miliQ water adjusted to pH 2.5 with 1 MHCl, prior to loading onto the cartridge; thus, the compounds pres-ent in this LMW fraction were at the same concentration as in thereal wine. The organic acids and sugars were removed with 5 ml ofwater adjusted to pH 2.5 with 1 M HCl (SPE water fraction or LMWfraction with acids). The polyphenols were eluted with 5 ml ofmethanol (SPE methanolic fraction or LMWW/OA fraction). Bothfractions were collected separately and evaporated to dryness.

2.4. Chemical analysis

2.4.1. Analysis of organic acids by UPLC–MSCompounds from the washing solvent fraction of SPE were ana-

lysed by UPLC–MS. Analyses were performed using a WatersAcquity Ultra Performance LCTM system (Waters, Milford, MA,USA) coupled to an ultraviolet (UV) detector (Waters, Milford,MA, USA) and a microTOF II high-resolution mass spectrometer(Bruker Daltonik, Bremen, Germany) equipped with an Apollo IIESI/APCI (electrospray/atmospheric pressure chemical ionization)multimode source. Bruker Daltonik (Version 3.4, Waters, Milford,MA, USA) was the software used to control the instrument andfor data acquisition and processing. UPLC separation was achievedusing an acquity BEH C18 (1.7 lm, 2.1 � 100 mm) column, pro-tected with a guard column with the same packing material(Waters, Milford, MA, USA) and kept at 40 �C. The solvent em-ployed was water/formic acid (0.1%) with an isocratic elution from0 to 6 min, washed with 100% acetonitrile for 2 min and equili-brated under the initial conditions for 2 min. Compounds wereeluted at a flow rate of 0.45 ml min�1 and the injection volume

was 7.5 ll. Samples were filtered through a 0.20 lm PTFE filter be-fore injection. Quantification was performed by UPLC-MS in nega-tive mode. Standard calibration curves were prepared for tartaric,citric, malic, lactic and succinic acid. Citramalic acid was quantifiedwith the citric acid standard and fumaric acid with the succinicacid standard. The electrospray source acquired between m/z 120and 1500 (capillary potential 3.5 kV, nebulizer gas 3.0 bar, dryinggas temperature 180 �C and flow 9.0 l min�1).

2.4.2. Analysis of phenolic compounds by UPLC–MSThe LMW fraction obtained from gel permeation chromatogra-

phy and the LMWW/OA fraction obtained from SPE were analysedby UPLC–UV and UPLC–MS. Each extract was analysed separatelyfollowing the methodology proposed by González, Avizcuri-Inac,Dizy, and Fernandez-Zurbano (2013). 7.5 ll of each sample dis-solved in water/formic acid 0.1%, previously filtered by 0.22 lmand at the same concentration of wine, was directly injected inthe UPLC system and chromatographed.

2.4.3. Standard wine analysisTotal polyphenol index (TPI) was estimated as the absorbance

at 280 nm multiplied by 100 (Ribereau-Gayon, 1970). The analysisof reducing sugars, ethanol content, pH and titratable and volatileacidities were determined by Infrared Spectrometry with FourierTransformation with a WineScan™ FT 120 (FOSS�, Barcelona,Spain), previously calibrated using official OIV methods.

2.5. Sensory analysis

2.5.1. Selection of sensory panelEighteen volunteers out of 34 (12 males and 6 females, aged be-

tween 21 and 45), all students or staff at the University of La Rioja(Spain), were selected for their ability to taste bitter on the basis oftheir PROP status (three supertasters and fifteen medium tasters)following the methodology described elsewhere (Gonzalo-Diagoet al., 2013). Panellists were not informed about the nature ofthe samples to be evaluated.

2.5.2. Panel trainingPanellists attended eight sensory training sessions (1 h per

session, twice a week) over 1 month. The training period includedtwo phases: a general phase and a product specific training phase.During the general training phase, different reference standardsolutions representative of taste, sucrose (0–12 g L�1) for sweet-ness, tartaric acid (0–1.5 g L�1) for acidity, quinine sulphate(0–10 mg L�1) for bitterness, and oral sensations, quercetin-3-O-galactoside (0–10 mg L�1) for velvety astringency (Scharbert,Holzmann, & Hofmann, 2004) and potassium and aluminiumsulphate (0–4 g L�1) for drying astringency stimuli (Sáenz-Navajas,Martín-López, Ferreira, & Fernández-Zurbano, 2011) were pre-sented to the panel to aid with recognition and discrimination.All standard solutions were dissolved in spring water. These attri-butes were rated on a ten-point scale (0 = ‘absence’, 1 = ‘very low’and 9 = ‘very high’) while persistence, also studied, was evaluatedon a nine-point scale (1 = ‘very short’ and 9 = ‘very long’). Thepersistence term was defined as the time in seconds that samplelinger after swallowing (5–10 s, short; 11–20 s, medium; 20–30 s,long). During wine specific training, different Spanish young redwines were presented to the judges. A single reference standardfor the astringency attribute (potassium and aluminium sulphate)was used to evaluate the wines. Two different specific types oftraining were carried out: one with LMW fractions; and anotherwith LMWW/OA fractions. For this purpose, different fractions ob-tained as described in Section 2.3 were presented and sweetness,acidity, bitterness, two different astringency sub-qualities, velvetyastringency, defined as a silky and finely textured kind of

190 A. Gonzalo-Diago et al. / Food Chemistry 154 (2014) 187–198

astringent sensation (notably perceived in the tip of the tongue andin front of superior teeth) and drying astringency defined as afeeling of desiccation or lack of lubrication (perceived in all partsof the mouth) and persistence were studied in these fractions.

2.5.3. Sample evaluationThe six wines (10 ml) and the six LMW fractions (4 ml) were de-

scribed in duplicate and in a single session, respectively. The sixLMWW/OA fractions (4 ml) were tasted and three were describedin duplicate in a single session. Both wines and fractions wereserved according to the protocol described by Gonzalo-Diagoet al. (2013). Samples were stored at 4 �C.

2.5.4. Sensory pilot studyFive trained judges participated in a sensory pilot study per-

formed with the aim of ascertain the sensory effectiveness of SPEmethod. This pilot study was carried out before being LMWW/OA

fractions sensory tasted by the eighteen trained panellists. A wineconsidered very astringent was selected and 4 ml of the differentfractions obtained as described in Section 2.3 were tasted (SPEwater fraction, SPE methanolic fraction and LMW fraction withoutSPE). Samples were described in duplicate in a single session andfollowing the protocol cited in Section 2.5.3.

2.6. Statistical analysis

A three-way ANOVA was used for each of the in-mouth attri-butes (taste, astringency and persistence) involved depending ofthe type of panel (wines, LMW fractions and LMWW/OA fractions),judge (J) and replicate (R) as fixed factors and all first order inter-actions were calculated to confirm panel performance using SPSSsoftware (IBM, Statistics version 19). Then, for attributes showinga significant wine-by-judge (W � J) interaction effect (P = 0.05), aPCA was run in order to assess judges’ disagreement.

For that purpose, a table with the wines in rows and the judges(mean scores of two repetitions) in columns was employed. A one-way ANOVA with repeated measurements was calculated on theattributes evaluated in-mouth – sweetness, acidity, bitterness,astringency and persistence – in order to evaluate the presenceof significant differences between the samples for these attributes.A one-way ANOVA with repeated measurements was also calcu-lated on the chemical variables correlated with bitterness, astrin-gency and persistence, in order to evaluate the presence ofsignificant differences between the samples for these variables.

Pearson’s correlation coefficients were calculated between sen-sory attributes and also between sensory and chemical parametersin order to evaluate significant correlations using Tukey’s test at(P < 0.05). The analyses were carried out using SPSS (Inc., Chicago,IL, USA) and SPAD software (version 5.5, CISIA-CESRESTA, Montre-uil, France).

Partial least-squares regression (PLSR) by means of the PLS1algorithm was used to evaluate the relationship between bitter-ness and non-volatile composition measured in the LMWW/OA frac-tions. Therefore, the matrix of chemical composition data wasscaled to obtain normalized data, and homoscedasticity and linear-ity were confirmed to avoid overestimating the goodness-of-fit incase data distribution would have been heteroscedastic. Finally,collinearity between the X-variables was ruled out because no sig-nificant correlations were found between variables. A full cross-validation was carried out to estimate the prediction ability ofthe models for a new set of samples. The data used to predict bit-terness with PLS were the average value given by all panellists toeach sample. PLS regression was performed with Unscrambler9.7 (CAMO, Oslo, Norway).

3. Results and discussion

3.1. Chemical and sensory characterisation of wines

The conventional oenological parameters of wines are shownin Table 1. All the wines presented normal values of ethanol(12.5–14.5%) with respect to other Spanish red wines. Reducingsugars contents were typical of dry wines, with values below4 g L�1. The pH’s ranged from 3.5 to 3.9, while the lowest titratableacidity value was observed for W6 (4.33 g L�1). Finally, W5 and W6wines had the highest total polyphenol index (TPI).

The ANOVA performed on the sensory data of the studied winesdetermined that the trained panellists were reproducible and con-sistent (data not shown). The results of the sensory evaluation car-ried out on the six wines studied are shown in Fig. 1. According tothis, the six wines were evaluated with high scores in acidity, bit-terness, astringency and persistence whereas the mean scores gi-ven for the sweetness attribute in all the wines were lower than2.5 on a ten point scale. It should be noted that neither sweetnessnor acidity showed significant differences. However, significantdifferences were observed in the case of bitterness, astringencyand persistence, with astringency being the attribute in which pan-ellists noted the greatest differences among the wines studied. Thecorrelation analysis between sensory attributes showed that bit-terness and astringency were positively correlated (r2 = 0.9456,P = 0.0011), and both terms were highly positively correlated withthe perception of persistence (p < 0.0010 in both cases). These datasuggest that persistence is mainly driven by oral sensations such asastringency and bitterness, as reported elsewhere (Sáenz-Navajas,Avizcuri et al., 2012; Sáenz-Navajas, Tao et al., 2010). Furthermore,these three terms - bitterness, astringency and persistence – werefound to be highly positively correlated with total polyphenol con-tent, TPI (P < 0.003 in all cases), suggesting that wine polyphenolcompounds contribute strongly to sensory perceptions, as reportedin previous studies (Arnold et al., 1980, Hufnagel & Hofmann,2008a, 2008b; Sáenz-Navajas, Avizcuri et al., 2012; Sáenz-Navajas,Tao et al., 2010).

3.2. Sensory characterisation of low molecular weight fractions(LMW-F)

The results of the ANOVA for the in-mouth assessed sensoryproperties of LMW fractions revealed that judges were a significant(P < 0.05) source of variation for all attributes since they have un-ique physiological perceptions (Bartoshuk, 1980). The judge effectis commonly found in sensory analyses and can be explained by in-ter-individual differences. The replicate effect was not significant,indicating a consistent assessment of replicates by the judges.The fraction-by-judge interaction (F*J) was not significant, indicat-ing that there were no differences in the interpretation of thisinteraction and that the judges were well trained with respect toall the attributes. The scores of the sensory evaluation carriedout on the six LMW fractions considered in the study are shownin Fig. 2a. As can be seen in this figure, all fractions were clearlyevaluated with high scores for acidity, this taste being the mostperceived among the samples. The scores given for acidity doubledthe scores given for astringency, bitterness or sweetness. This highscore given for acidity may have been due to the presence in theseLMW fractions of non-volatile organic acids such as tartaric acidsand lactic acids commonly present in wines which have undergonemalolactic fermentation. As shown in Table 2, tartaric acid, lacticacid and succinic acid were found at higher concentrations thantheir sensory threshold for acidity; thus, these acids are likely tohave an impact on wine acidity. Similar findings were also reportedby Hufnagel and Hofmann (2008a), who highlighted the presence

Table 1Denomination of Origin or region, vintage, varietal composition, conventional analysis and total polyphenol index (TPI) of the six wines selected for this study.

Denomination of Origin or region Vintage Grape variety Ethanol (v/v) pH Volatile aciditya Total aciditya Reducing sugarsb TPIc

W1 D.O.Ca. Rioja 2008 Tempranillo 12.5 3.62 0.32 4.81 1.18 30.0W2 VT Cangas 2008 Mencía. Albarín negro 12 3.47 0.31 5.05 1.22 36.8W3 D.O. Valencia 2008 Bobal. Shyrah 12.5 3.54 0.30 5.25 3.98 40.8W4 VT Castilla y León 2007 Tempranillo 13 3.57 0.31 5.14 1.66 55.3W5 D.O. Valdepeñas 2008 Tempranillo 13.5 3.73 0.26 5.33 1.59 59.0W6 D.O. Toro 2008 Tempranillo 14.5 3.89 0.36 4.33 1.69 62.5

a Expressed as g L�1 of tartaric acid.b Expressed as g L�1.c Total polyphenol index expressed as absorbance at 280 � 100.

WINES

0

1

2

3

4

5

6

7

Sweet Acid Bitter Astringency Persistence

Mea

n sc

ores

W1W2W3W4W5W6

nds

nds

c

aa

bc bcb

a

d

c

b

d

c

a

ab ab

bbb

nds

nds

c

aa

bc bcb

a

d

c

b

d

c

a

ab ab

bbb

Fig. 1. Mean sensory ratings for the six wines. Error bars are calculated as s/(n)1/2; (s) standard deviation; (n) number of panellists. Different letters indicate the existence of asignificant difference between wines (a 6 0.05) (Tukey’s test); nds, no significant differences.

A. Gonzalo-Diago et al. / Food Chemistry 154 (2014) 187–198 191

of L-tartaric acid as a very important stimulus for acid taste. Thegreat acidity perceived may be also explained by the absence inthese LMW fractions of a high molecular weight fraction consid-ered and evaluated as astringent in a previous study (Gonzalo-Diagoet al., 2013) and which may have modulated the perception of acid-ity, increasing it with respect to wines. This acidity-astringencyinteraction is consistent with the findings of Frank, Wollmann,Schieberle, and Hofmann (2011) in a taste omission study in whichthe lack of the high molecular weight fraction increased the inten-sity of sourness. In spite of the high acidity of LMW fraction, thepanellists described it as astringent and bitter. However, no signif-icant conclusions could be drawn from these experimental data,except that the high acidity perceived in LMW fractions may havemasked the other perceptions. It is important to highlight thatthere was no relationship between the scores given to the sensoryattributes evaluated in the wines and those given to the fractions.

3.3. Elimination of acids

3.3.1. A sensory pilot studyAn exploratory sensory pilot study was conducted to evaluate

the sensory effectiveness of the SPE method to retain the phenoliccompounds and to remove the organic acids, in order to eliminateinterferences caused by the high acidity perceived in LMW frac-tions. That’s why five trained judges participated in this pilot sen-sory session carried out before being LMWW/OA fractions sensorytasted by the eighteen trained panellists. In it, the SPE water frac-tion, the SPE methanolic fraction and the LMW fraction withoutSPE from a single very astringent wine were evaluated. As regards

this sensory study, the SPE water fraction or fraction with acidswas evaluated as clearly acidic and persistent with low values forbitterness and astringency (Supplementary data). In this case, thepersistence perceived by the panellists may have been due to theacidity perceived. In contrast, the SPE methanolic fraction or frac-tion without acids scored with values of 1 out of 9 for acid taste,retained the bitterness and astringency of the original sample(LMW-F). In this case, when the acidity was barely perceived, thepersistence perceived may have been due to the astringency andbitterness.

3.3.2. Elimination of acids from LMW-FOnce panellists had not perceived acid taste in the fraction

without acids, the next step was to check the effectiveness of theSPE method for retaining the phenolic compounds. The chromato-grams of the fractions resulting from the application of the SPEprocess to one of the samples (Supplementary data) showed thatno compounds appeared over the 2 min of elapsed time in the frac-tion corresponding to the washing solvent step. This fraction wasmainly composed of organic acids such as tartaric, lactic, citrama-lic, succinic and cis- and trans-aconitic acids and sugars. In thechromatogram corresponding to the fraction eluted with metha-nol, all compounds appeared over the 2 min of elapsed time. Theresults showed that the recovery of certain compounds in themethanolic fraction was low, e.g., the gallic and protocatechuicacids were especially eluted in the water fraction at 90% and57%, respectively. Some hydroxycinnamic acids, such as caftaricacid and coutaric acid, were also eluted in the water fraction butin a smaller proportion of 7% and 0.5%, respectively. Catechin

-1

0

1

2

3

4

5

6

7

Sweet Acid Bitter VelvetyAstringency

DryingAstringency

Persistence

Mea

n sc

ores

W/OA-F W1

W/OA-F W2

W/OA-F W3

W/OA-F W4

W/OA-F W5

W/OA-F W6

LMW-WITHOUT ACIDS

a

a

a

bc

b

ndsc

c

b

c

bc

nds

abnds

a

bc

bcbc

bb

c

0

1

2

3

4

5

6

7

8

Sweet Acid Bitter VelvetyAstringency

DryingAstringency

Persistence

Mea

n sc

ores

LMW-F W1

LMW-F W2

LMW-F W3

LMW-F W4

LMW-F W5

LMW-F W6

aa

a

a

b

a

nds

a

a

b

ab

b

nds

nds

nds

ab

LMW-FRACTION(a)

(b)

Fig. 2. (a) Mean sensory ratings for the six wine fractions containing the low molecular weight compounds. (b) Mean sensory ratings of the six wine fractions containing thelow molecular weight compounds without acids. Error bars are calculated as s/(n)1/2; (s) standard deviation; (n) number of panellists. Different letters indicate the existenceof a significant difference between wines (a 6 0.05) (Tukey’s test); nds, no significant differences.

192 A. Gonzalo-Diago et al. / Food Chemistry 154 (2014) 187–198

was the only phenolic compound tested that eluted in the waterfraction in a proportion of approximately 3%. These recoveryresults coincide with those reported by Pérez-Magariño,Ortega-Heras, and Cano-Mozo (2008) in a study focused on theisolation of phenolic compounds in red wines. These authorsalso found that the most polar compounds such as gallic and pro-tocatechuic acid appeared mainly in the water fraction and trans-caftaric acid was distributed in both fractions whereas trans-coutaric acid and (+)-catechin appeared in traces in the waterfraction. Sun, Leandro, de Freitas, and Spranger (2006) have alsodescribed that alcohol reduces the retention of certain phenoliccompounds by the sorbent during the loading and washing phases,especially hydroxybenzoic acids. In this study, samples were deal-coholized, but the loss of hydroxybenzoic acids was confirmed.

3.4. Sensory evaluation of fractions without acids (LMWW/OA-F) fromsix red wines

A three-way ANOVA for each of the in-mouth attributes (sweet-ness, acidity, bitterness, velvety and drying astringency and persis-tence) involving fractions, judge and three replicates as fixed

factors were calculated. The judge effect was significant on allattributes. However, the replicate effect was not significant, indi-cating a consistent assessment of attributes and reflecting thereproducibility of the panel. Fraction-by-judge (F � J) interactionwas significant only for drying astringency. A PCA run on dryingastringency attribute revealed that only one judge was causingthe differences in the interpretation of this term. Thus, the scoresgiven by this judge for drying astringency attribute were not con-sidered. Fig. 2b shows the sensory evaluation of the six low molec-ular weight fractions without acids (LMWW/OA-F). As shown in thisfigure and as was expected, acidity and sweetness were rated withvery low values (<1). LMWWO/A fractions were described as bitter,persistent and also slightly astringent. In terms of astringency, allfractions were described with the velvety sub-quality althoughno significant differences were reported between them. However,the panellists only described two samples, W5 and W6, with theattribute drying astringency, in the case of wines presenting higherTPI (Table 1). The other fractions were described with very lowscores for this term (<0.4). The values given to velvety astringencyin the fractions without acids (LMWW/OA-F) were higher than thosegiven for LMW fractions. In contrast, the scores assigned to drying

Table 2Compounds analysed by UPLC–DAD–MS in both fractions obtained from SPE (water fraction and methanolic fraction). Average, maximum (max) and minimum (min)concentrations, max/min, taste threshold (expressed in mg L�1) and Dose-over-Threshold (DoT) factors for acidity, bitterness and astringency. Significant compounds obtained bybivariate Pearson’s correlation are marked in bold. (Listed in decreasing concentration within each group).

Compounds Average(mg L�1)

Max(mg L�1)

Min(mg L�1)

Max/min

Taste thresholde

(mg L�1)DoTmax

a

Acids derivativesLactic acidf 1200 2370 320 7.04 1393b 1.7b

Tartaric acidf 730 1020 Tr 44b 23.3b

Succinic acidf 170 340 Tr 106b 3.2b

Citramalic acidf 50 130 Tr n.aCoutaric acidh 41.0 64.2 29.2 2.20 10.00c 6.4c

trans-caftaric acid h 30.7 37.5 26.2 1.43 5.00c 7.5c

Caffeic acidg 7.14 11.1 2.88 3.84 13.00c 0.8c

trans-aconiticf 3.78 6.44 2.16 2.98 87b/0.09c <0.1b/71.5c

Caffeic acid ethyl esterg 2.50 5.60 1.32 4.24 58c/229d <0.1c/<0.1d

Ferulic acidg 2.54 3.19 1.20 2.65 13.00c <0.1c

Gallic acidh 1.57 2.23 0.90 2.48 50.00c <0.1c

Syringic acidg 1.88 2.22 1.31 1.70 52c <0.1c

Vanillic acidg 1.28 2.02 0.61 3.32 53c <0.1c

cis-aconiticf 1.37 1.64 1.05 1.56 87b/0.09c <0.1b/18.2c

cis-coumaric acidg 0.99 1.40 0.28 4.94 n.aProtocatechuic acidh 0.47 0.62 0.24 2.63 32.00c <0.1c

trans-coumaric acidg 0.23 0.29 0.21 1.36 23.00c <0.1c

Protocatechuic acid ethyl esterg 0.13 0.19 0.08 2.40 9c/182d <0.1c/<0.1d

Citric acidf nd nd nd 499b

Fumaric acidf nd nd nd n.aMalic acidf nd nd nd 494b

Flavonolsg

Myricetin 3.88 9.25 3.50 2.64 10.0c/10.0d 0.9c/0.9d

Quercetin 3.85 9.36 0.73 12.87 10.0c/10.0d 0.9c/0.9d

Quercetin-3-O-glucuronide 3.24 7.40 5.90 1.25 1.00c 7.4c

Myricetin-3-O-glucoside 3.41 6.80 2.33 2.92 1.0c 6.8c

Quercetin-3-O-galactoside 1.00 4.06 0.46 8.78 0.20c 20.3c

Syringetin-3-O-galactoside 0.41 0.84 0.24 3.49 n.aIsorhamnetin-3-O-glucoside 0.14 0.59 0.10 6.20 1.20c <0.1c

Quercetin-3-O-rutinoside 0.22 0.48 0.03 18.72 0.0006c 685c

Quercetin-3-O-glucoside 0.05 0.21 0.10 2.10 1.00c 0.2c

Kaempferol 0.07 0.20 nd 20.0c/20.0d <0.1c/<0.1Kaempferol-3-O-glucoside nd nd nd nd 0.3c

Isorhamnetin nd nd nd nd

Flavanolsg

Catechinh 34.30 64.18 14.39 4.46 119c/290d 0.5c/0.2d

Procyanidin B2 17.84 38.08 5.77 6.60 110c/280d 0.3c/0.1d

Procyanidin B1 11.06 25.72 2.58 9.96 139c/231d 0.2c/<0.1d

Epicatechin 13.12 22.96 4.47 5.13 270c/270d <0.1c/<0.1d

Epigallocatechin 7.23 15.55 1.94 7.99 159c <0.1c

Procyanidin A2 1.16 3.34 nd n.aEpicatechin gallate 0.06 0.38 nd 115c <0.1c

Catechin gallate 0.15 0.31 nd 110c <0.1c

Procyanidin C1 0.08 0.19 nd 260c/347d <0.1c/<0.1d

Anthocyaninsg

Malvidin-3-O-glucoside 46.96 80.55 26.03 3.09 n.aDelphinidin-3-O-glucoside + Malvidin-3-O-glucoside-(epi)catechin 12.26 29.79 4.60 6.47 n.aPeonidin-3-O-glucoside 14.31 26.35 6.31 4.17 n.aPetunidin-3-O-glucoside 11.05 26.12 5.22 5.00 n.aMalvidin-3-O-glucoside-8-ethyl-(epi)catechin + Petunidin-3-O-

(60-acetyl)-glucoside2.65 5.16 0.91 5.67 n.a

Malvidin-3-O-glucoside-4-vinylcatechol + Malvidin-3-O-(60-p-coumaroyl)-glucoside trans

1.54 4.33 nd n.a

Petunidin-3-O-(60-p-coumaroyl)-glucoside 1.04 3.51 0.26 13.31 n.aCyanidin-3-O-glucoside 1.65 3.26 0.70 4.64 n.aA-type vitisin of malvidin-3-O-glucoside 1.64 2.58 0.91 2.82 n.aDelphinidin-3-O-(60-p-coumaroyl)-glucoside 0.93 2.01 0.43 4.63 n.aDelphinidin-3-O-(60-acetyl)-glucoside 1.00 1.96 0.58 3.37 n.aMalvidin-3-O-(60-acetyl)-glucoside 0.23 0.87 nd n.aA-type vitisin of malvidin-3-O-(60-p coumaroyl)-glucoside 0.51 0.86 0.21 4.07 n.aB-type vitisin of malvidin-3-O-glucoside 0.39 0.73 0.24 3.09 n.aB-type visitin of delphinidin-3-O-glucoside 0.38 0.71 0.17 4.24 n.aPeonidin-3-O-(60-p-coumaroyl)-glucoside 0.41 0.68 0.18 3.85 n.aMalvidin-3-O-glucoside-4-vinylphenol 0.27 0.57 nd n.aA-type vitisin of peonidin-3-O-glucoside 0.27 0.49 nd n.a

(continued on next page)

A. Gonzalo-Diago et al. / Food Chemistry 154 (2014) 187–198 193

Table 2 (continued)

Compounds Average(mg L�1)

Max(mg L�1)

Min(mg L�1)

Max/min

Taste thresholde

(mg L�1)DoTmax

a

Cyanidin-3-O-(60-acetyl)-glucoside 0.24 0.41 nd n.acis-malvidin-3-O-(60-p-coumaroyl)-glucoside 0.14 0.28 nd n.aB-type vitisin of malvidin-3-O-(60-acetyl)-glucoside 0.09 0.22 nd n.aMalvidin-3-O-(60-p-coumaroyl)-glucoside-ethyl-(epi)catechin 0.05 0.17 nd n.a

nd: not detected; n.a: not available; Tr: traces.a The DoTmax is calculated as the ratio of the concentration and taste threshold.b Taste threshold or DoT factor for acidity.c Taste threshold or DoT factor for astringency.d Taste threshold or DoT factor for bitterness.e Sensory thresholds have been taking from Hufnagel et al. (2008a); Scharbert et al. (2004); Dadic and Belleau (1973) and Kamura and Watanabe (1981).f Compounds identified in SPE water fraction.g Compounds identified in SPE methanolic fraction.h Compounds identified in both SPE fractions, water and methanolic.

194 A. Gonzalo-Diago et al. / Food Chemistry 154 (2014) 187–198

astringency in LMWW/OA fractions were lower than those given forLMW fractions. This may have been due to the absence of organicacids such as tartaric, succinic, lactic and cis and trans aconiticacids because both, cis and trans aconitic acids, have been de-scribed as astringent (Hufnagel & Hofmann, 2008a; Sáenz-Navajas,Avizcuri et al., 2012a). Other phenolic compounds that could haveelicited astringency and that may have been partially eliminated inthis LMWW/OA fractions were gallic acid, protocatechuic acid, trans-caftaric acid, coutaric acid and catechin. These cited compoundsmay also have been responsible for the decrease in drying astrin-gency perceived because all of them were described with the attri-bute puckering astringency (Hufnagel & Hofmann, 2008a).

In terms of bitter taste, the results showed that fractions with-out acids were significantly different among them, these differ-ences being greater than the differences perceived in LMWfractions (Fig. 2a). It should be noted that the bitterness scores ofLMWW/OA fractions were not correlated with either the sensory bit-terness scores of LMW fractions or the sensory bitterness scores ofwines. There are various explanations for this finding. Alcohol con-centration present in wines and not in the fractions has been re-ported to increase the perception of bitterness (Nurgel &Pickering, 2006). The lack of ethanol in the fractions may have ac-counted for the lower scores given for bitterness in the fractionsfrom wines W1, W2, W3 and W4. However, the same cannot besaid of the fractions from W5 to W6 because both fractions wereevaluated with scores next to wines (Fig. 3). Moreover, thecontribution of organic acids to bitter taste does not seem to havebeen relevant. The elimination of organic acids described as bitter

BITTERNESS

0

1

2

3

4

5

6

W/F1 W/F2 W/F3 W/F4

Mea

n sc

ores

b

a

b

b

a

b

b

a

b b

a

b

0

1

2

3

4

5

6

W/F1 W/F2 W/F3 W/F4

Mea

n sc

ores

b

a

b

b

a

b

b

a

b b

a

b

Fig. 3. Comparison of the bitterness scores given to both wines and the two fractionsdeviation; (n) number of panellists. Different letters indicate the existence of a significa

(Whiting, 1976) did not reduce the bitterness perceived in the frac-tions because all fractions without acids were described as morebitter, except fraction 3. As mentioned previously, special attentionmust be drawn to the differences observed in bitterness betweenthe W5 and W6 fractions without acids and the other four fractions(Fig. 3). The low molecular weight compounds present in LMWW/OA

fractions were perceived as bitter in low concentrations and bitterand astringent in higher concentrations. These findings are consis-tent with other research describing several low weight phenoliccompounds as both bitter and astringent (Hufnagel & Hofmann,2008a; Kallithraka et al., 1997). Additionally, it is possible that amasking effect occurred between both sensory attributes. Thus,when the compounds were present in low concentrations, in theW1, W2, W3 and W4 fractions, only the bitter taste was perceived.However, when concentrations were higher, as in the W5 and W6fractions, astringency could be perceived but also the noticeableperception of bitter taste increased, with similar values to wines.In this case, and as reported by Stevens (1997) bitter taste wouldbe the predominant quality when perceived astringency was low,being possible a synergetic effect when a low astringency was per-ceived as in the case of the W5 and W6 fractions. This effect couldbe extrapolated to wines (Fig. 1) where the less astringent wines(W1 and W2) were evaluated more bitter than astringent.However, a contrary pattern may happen in the case of the mostastringent wines (W5 and W6), which were evaluated with lowerscores for bitterness than for astringency. This bitter taste sensorymodulation due to the different astringency perception may haveaccounted for the small differences detected in the bitter taste of

W/F5 W/F6

WineLMW fraction LMW w/oa fraction

b

a

a

b

a

a

W/F5 W/F6

WineLMW fraction LMW w/oa fraction

b

a

a

b

a

a

studied, LMW-F and LMWW/OA-F. Error bars are calculated as s/(n)1/2; (s) standardnt difference between samples (a 6 0.05) (Tukey’s test).

A. Gonzalo-Diago et al. / Food Chemistry 154 (2014) 187–198 195

wines. It is important to note that all the panellists were able to de-tect bitter taste in the samples. Further research would be neces-sary to corroborate this bitterness–astringency interaction.

In terms of persistence, the elimination of acids allowed thepanellists to perceive significant differences among fractions, dif-ferences not observed in the presence of acids (Fig. 2b and a,respectively). A significantly positive correlation (F = 23.78,P = 0.0082, r2 = 0.86) was found between bitterness andpersistence.

3.5. Sensory in-mouth properties of LMWW/OA fraction and theircorrelation with its chemical composition

3.5.1. BitternessBitter taste from LMWW/OA fractions was positively correlated

with coutaric acid, procyanidin B1, catechin gallate and kaempher-ol (P < 0.02), with (+)-catechin, procyanidin B2, malvidin-3-O-glu-coside-8-ethyl epicatechin + petunidin-3-O-(6-acetyl)-glucoside,A-type vitisin of malvidin-3-O-6-p-coumaroyl-glucoside, cis-mal-vidin-3-O-6-p-coumaroyl-glucoside (P < 0.05) and with quercetin-3-O-rutinoside, petunidin-3-O-6-p-coumaroyl-glucoside (P < 0.08)and negatively with protocatechuic acid ethyl ester (P < 0.08)(Supplementary data). In general, the concentrations found forall the above compounds were below their sensory thresholds(Table 2). These low concentrations make difficult to predictwhether these compounds and the other correlated compoundsfor which sensory thresholds were not identified are capable ofeliciting bitter taste. However, these findings are consistent withthose reported by Scharbert and Hofmann (2005) who found thatthe omission of flavonol-3-glycosides and flavanols lowered theperceived bitterness, both flavonoids being the key compoundsin the taste of black tea. In this connection, Laaksonen, Sandell,and Kallio (2010) pointed out that certain derivatives of hydroxy-cinnamic acids may contribute to bitterness while Preys et al.(2006) described certain flavonol aglycones, myricetin and querce-tin as bitter compounds. Furthermore, other authors have pointedout that wine bitterness may be induced by sub-threshold concen-trations of flavanols and phenolic acid ethyl esters (Hufnagel &Hofmann, 2008a). Consequently, in this study, two flavanols(catechin and catechin gallate), two dimeric phenols (procyanidinB1 and B2), a hydroxycinnamic acid (coutaric acid), two flavonols(kaempherol and quercetin-3-O-rutinoside) and four anthocyaninswere relevant compounds in the bitterness evaluated in LMWW/OA

fractions. To the best of our knowledge, this is the first time thatcoutaric acid, kaempherol and quercetin-3-O-rutinoside have beensensory described as bitter potentially compounds and also thefirst time that protocatechuic acid ethyl ester, despite being clas-sified as a bitter compound, has been negatively correlated withbitter taste. Until now, quercetin-3-O-rutinoside has beendescribed as a non-bitter astringent taste. However, it has beenshown that the presence of quercetin-3-O-rutinoside increasesthe bitterness intensity perceived for other bitter compounds suchas caffeine (Scharbert & Hofmann, 2005).

3.5.1.1. Bitterness prediction from non-volatile composition by PLSR.PLSR has been successfully used to study the relationships be-tween sensory and chemical analyses in wine. Thus, to build apredictive model for bitterness (Y variable), using non-volatilecomposition (X variable) as an explicative variable, PLSR withcross-validation was performed. At first approximation, in orderto reduce the number of compounds that could have beenresponsible for sensory modifications, only the compounds witha significant Pearson’s correlation were considered due to thelack of the bitterness sensory threshold for many of the com-pounds analysed. Furthermore, it was to be expected that the

concentrations of the taste-active compounds responsible foreffective sensory differences in the studied samples displayedsubstantial differences. Thus, the maximum/minimum concentra-tion rate was taken as a differentiability criterion (Escudero,Campo, Fariña, Cacho, & Ferreira, 2007). Compounds reachingvalues above 2 for max/min parameter were considered to havethe greatest capacity to induce sensory modifications and wereconfirmed by calculating significant differences by means ofANOVA. As shown in Table 2, the most discriminant taste-activecompounds, on the basis of the quotient maximum/minimum,were found to be quercetin-3-O-rutinoside (max/min = 18.72),petunidin-3-O-6-p-coumaroyl-glucoside (max/min = 13.31), procy-anidin B1, (max/min = 9.96), procyanidin B2, (max/min = 6.60), mal-vidin-3-O-glucoside-8-ethyl epicatechin + petunidin-3-O-(6-acetyl)-glucoside (max/min = 5.67) and catechin (max/min = 4.46). Asecond group involved compounds with a maximum/minimumratio ranging between 4 and 2 comprising A-type vitisin of malvi-din-3-O-6-p-coumaroyl-glucoside, coutaric acid and protocatechuicacid ethyl ester.

As shown in Fig. 4a and b, the total variance explained by thefirst principal component was 86% (82% by cross-validation) andthe root-mean-square prediction error (RMSEP) was 0.588.

The model included nine significant variables: quercetin-3-O-rutinoside (Q-rut), petunidin-3-O-6-p-coumaroyl-glucoside(Pet-cou-glu), procyanidin B1 (PC-B1), procyanidin B2 (PC-B2),malvidin-3-O-glucoside-8-ethyl epicatechin + petunidin-3-O-(6-acetyl)-glucoside (Ant-flav + Pet-ace-glu), catechin (Cat), A-typevitisin of malvidin-3-O-6-p-coumaroyl-glucoside (Mv-glu-pyr),coutaric acid (Cout) and protocatechuic acid ethyl ester (Procat).All variables were positively correlated to bitterness except proto-catechuic acid ethyl ester and the importance of these variables isshown in the predicted model: Bitterness = 1.036 + 0.014 ⁄ Cout� 3.928 ⁄ Procat + 0.020 ⁄ PCB1 + 0.009 ⁄ Cat + 0.015 ⁄ PCB2 + 1.273 ⁄

Q-rut + 0.126 ⁄ Ant-flav + Pet-ace-glu + 0.656 ⁄Mv-glu-pyr + 0.155 ⁄

Pet-cou-gluThe compound that showed a greater weight in the model

was the phenolic acid protocatechuic acid ethyl ester. To ourknowledge, this is the first time that three anthocyanins havebeen related to bitter taste, A-type vitisin of malvidin-3-O-6-p-coumaroyl-glucoside being the anthocyanin that elicited thegreatest contribution in the model. The flavonol quercetin-3-O-rutinoside also appeared to have a strong contribution, whereasthe flavanol monomer catechin and the dimers made a slightcontribution; the trimer C1, also found in these samples, wasnot correlated with bitter taste. Thus, Hufnagel and Hofmann,(2008b) by means of test dilution analysis observed that winefractions identified as bitter contained hydroxycinnamic andhydroxybenzoic acids ethyl esters while wine fractions present-ing monomers and flavanol dimers were not detected as bitter.This finding seems agree with the results obtained in this study,in relation to the low contribution observed for catechin andprocyanidins B1 and B2 in the predicted model, although as pre-vious researches have reported, flavanol monomers and dimersare important taste compounds in red wines (Arnold et al.,1980). Weber, Greve, Durner, Fischer, and Winterhalter (2013)studying the sensory and chemical characterisation of phenolicpolymers from red wine obtained by gel permeation chromatog-raphy did not find any correlation between the flavanols andanthocyanins studied and bitter taste while Soares et al. (2013)found that different phenolic compounds as epicatechin, atrimmer procyanidin and malvidin-3-O-glucoside activateddistinct human bitter taste receptors.

3.5.2. AstringencyThe flavon-3-ol glucosides have been described using the

term velvety astringency (Hufnagel & Hofmann, 2008a), but no

Fig. 4. PLS regression: (a) Plot of predicted vs. measured bitterness scores obtained with the (a) calibration (in red) and (b) validation (in blue), (b) X- and Y-loading plot ofbitterness prediction by PLS regression model. Significant X-variables and predicted variable (bitterness) are plotted on the first two principal components (PC1 and PC2). (Forinterpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

196 A. Gonzalo-Diago et al. / Food Chemistry 154 (2014) 187–198

sensory significant differences were observed among LMWW/OA

for velvety sensation. Conversely, drying astringency was usefulfor characterising differences among samples, allowing chemicaldata and sensory scores given for this term to be correlated.The compounds found to be positively correlated with dryingastringency were: coutaric acid, (+)-catechin, procyanidin B1,procyanidin B2, kaempherol and A-type vitisin of malvidin-3-O-(60-p-coumaroyl)-glucoside at a P < 0.02, (�)-epicatechin, (�)-epi-gallocatechin, procyanidin C1, quercetin-3-O-rutinoside, querce-tin, malvidin-3-O-glucoside-8-ethyl epicatechin + petunidin-3-O-(60acetyl)-glucoside, petunidin-3-O-(60-p-coumaroyl)-glucoside atP < 0.05 and (�)-catechin gallate and malvidin-3-O-glucoside atP < 0.1 (Supplementary data). In terms of the sensory thresholds(Table 2), of all the compounds described above, only coutaricacid and quercetin-3-O-rutinoside presented concentrationsabove its sensory thresholds. For this reason, both compoundswere determined to be the most responsible for eliciting dryingastringency in LMWW/OA fractions. Quercetin-3-O-rutinoside,along with other quercetin glycosides, is one of the key astrin-gent compounds in black tea (Scharbert et al., 2004), being ableto elicit astringent properties at extremely low concentrations.Similar to findings described for bitter taste, Scharbert andHofmann (2005) showed that the omission of flavonol glycosidessignificantly affected the perception of velvety and drying

astringency. Laaksonen et al. (2010) found that phenolic com-pounds, especially flavonol glycosides and hydroxycinnamic acidderivatives, correlated strongly with the astringency of berryfractions. The other compound, coutaric acid, has been proposedin different astringency predictive models built with two differ-ent groups of wines (Sáenz-Navajas, Avizcuri et al., 2012;Sáenz-Navajas, Tao et al., 2010). In this study, four anthocyaninswere found to be correlated with perceived astringency. Oberhol-ster et al. (2009) observed that the presence of anthocyaninsduring fermentation increased the intensity of astringency attri-butes; the presence of anthocyanins in red wine may accountfor the differences perceived between the mouthfeel propertiesof a white and a red wine. The addition of anthocyanins mainlycontributed to the fine grain sub-attribute of astringency(Oberholster et al., 2009). Brossaud, Cheynier, and Noble (2001)found that the addition of an anthocyanin mixture to seed andskin tannin extracts increased the astringency of the solutionover the astringency of either fraction alone but had no effecton bitterness. Another study found that an anthocyanin fractionincreased the perceived astringency and ‘fullness’ of a modelwine however, this increase was relatively unimportant becausethe score given to this anthocyanin fraction was lower than thescore given to the least astringent proanthocyanidin fraction(Vidal, Francis, Williams et al. 2004).

A. Gonzalo-Diago et al. / Food Chemistry 154 (2014) 187–198 197

Thus, it seems that certain low molecular weight compoundscontributed to the astringency perceived in the fractions fromwines W5 to W6. Nevertheless, further sensory research need tobe done bringing together different concentrations of the signifi-cant compounds involved in both attributes, bitterness and astrin-gency, in order to extend the knowledge in this sensory field.

3.5.3. PersistencePersistence evaluated in the LMWW/OA fractions was correlated

with quercetin-3-O-rutinoside and petunidin-3-O-(60-p-couma-royl)-glucoside at (P < 0.02), with coutaric acid at (P < 0.05) andwith trans-coumaric acid, procyanidin B1, procyanidin B2, (+)-cat-echin, (�)-catechin-gallate and kaempherol at (P < 0.08) (Supple-mentary data). All these cited compounds, with the exception oftrans-coumaric acid, were also correlated with bitterness andastringency attributes. This observation was expected because, asNoble (1990) stated, bitterness and astringency attributes are char-acterised by long persistence. Although in this study trans-couma-ric acid was not correlated with perceived astringency, trans-coumaric acid has been described as an astringent compound(Hufnagel & Hofmann, 2008a).

4. Conclusions

In summary, this study has provided additional evidence withrespect to the chemical basis of bitter taste in wines. The elimina-tion of astringent and acid compounds allowed to build a statisti-cally significant model to explain bitter taste evaluated in thesesamples by panellists selected for their ability to detect bitter taste.Furthermore, a hypothesis regarding bitterness–astringency inter-action has been proposed. For that reason, this paper has drawnkey conclusions that will be helpful for further research into winesand sensory analysis related to bitter taste. Further research is re-quired in order to validate the sensory contribution of compoundsincluded in the predicted model, as well as to verify the sensoryinteractions observed among compounds.

Acknowledgements

The authors would like to thank the judges participating in thesensory panel for their great contribution to this research. This re-search was supported by funds from the Ministry of Education andScience under the National Scientific R&D&I Plan (Pro-ject:AGL2010-22355-C02-02). The author A.G.D. receives supportfrom the University of La Rioja’s FPI Research FellowshipProgramme.

Appendix A. Supplementary data

Supplementary data associated with this article can be found, inthe online version, at http://dx.doi.org/10.1016/j.foodchem.2013.12.096.

References

Arnold, R. A., Noble, A. C., & Singleton, V. L. (1980). Bitterness and astringency ofphenolic fractions in wine. Journal of Agricultural and Food Chemistry, 28(3),675–678.

Bartoshuk, L. M. (1980). Separate worlds of taste. Psychology Today, 14, 48.Brossaud, F., Cheynier, V., & Noble, A. C. (2001). Bitterness and astringency of grape

and wine polyphenols. Australian Journal of Grape and Wine Research, 7, 33–39.Chira, K., Pacella, N., Jourdes, M., & Teissedre, P. L. (2011). Chemical and sensory

evaluation of Bordeaux wines (Cabernet-Sauvignon and Merlot) and correlationwith wine age. Food Chemistry, 126(4), 1971–1977.

Dadic, M., & Belleau, G. (1973). Polyphenols and beer flavour. Proceedings of theAmerican Society of Brewing Chemists, 4, 107–114.

de Villiers, A., Lynen, F., Crouch, A., & Sandra, P. (2004). Development of a solid-phase extraction procedure for the simultaneous determination of polyphenols,organic acids and sugars in wine. Chromatographia, 59, 403–409.

Escudero, A., Campo, E., Fariña, L., Cacho, J., & Ferreira, V. (2007). Analyticalcharacterization of the aroma of five premium red wines. Insights into the roleof odor families and the concept of fruitiness of wines. Journal of Agricultural andFood Chemistry, 55(11), 4501–4510.

Fischer, U., & Noble, A. C. (1994). The effect of ethanol, catechin concentration, andpH on sourness and bitterness of wine. American Journal of Enology andViticulture, 45, 6–10.

Frank, S., Wollmann, N., Schieberle, P., & Hofmann, T. (2011). Reconstitution of theflavor signature of Dornfelder red wine on the basis of the naturalconcentrations of its key aroma and taste compounds. Journal of Agriculturaland Food Chemistry, 59, 8866–8874.

Gawel, R., Francis, L., & Waters, E. J. (2007). Statistical correlations between the in-mouth textural characteristics and the chemical composition of Shiraz wines.Journal of Agricultural and Food Chemistry, 55(7), 2683–2687.

González, M., Avizcuri-Inac, J. M., Dizy, M., & Fernandez-Zurbano, P. (2013).Rapid screening for polyphenols and organic acids in red wine by ultraperformance liquid chromatographic coupled to ultraviolet–vis and massspectrometry detectors. In High-performance liquid chromatography (HPLC):Principles, practices and procedures. Hauppauge, New York: Nova SciencePublishers (in press).

Gonzalo-Diago, A., Dizy, M., & Fernández-Zurbano, P. (2013). Taste andmouthfeel properties of red wines proanthocyanidins and their relation tothe chemical composition. Journal of Agricultural and Food Chemistry, 61,8861–8870.

Hufnagel, J. C., & Hofmann, T. (2008a). Quantitative reconstruction of thenonvolatile sensometabolome of a red wine. Journal of Agricultural and FoodChemistry, 56(19), 9190–9199.

Hufnagel, J. C., & Hofmann, T. (2008b). Orosensory-directed identification ofastringent mouthfeel and bitter-tasting compounds in red wine. Journal ofAgricultural and Food Chemistry, 56(4), 1376–1386.

Kallithraka, S., Bakker, J., & Clifford, M. N. (1997). Evaluation of bitterness andastringency of (+)-catechin and (�)-epicatechin in red wine and in modelsolution. Journal of Sensory Studies, 12(1), 25–37.

Kamura, S., & Watanabe, M. (1981). Determination of phenolic cinnamates in whitewine and their effect on wine quality. Agricultural and Biological Chemistry, 45,2063–2070.

Laaksonen, O., Sandell, M., & Kallio, H. (2010). Chemical factors contributing toorosensory profiles of bilberry (Vaccinium myrtillus) fractions. European FoodResearch and Technology, 231, 271–285.

Landon, J. L., Weller, K., Harbertson, J. F., & Ross, C. F. (2008). Chemical and sensoryevaluation of astringency in Washington State red wines. American Journal ofEnology and Viticulture, 59(2), 153–158.

Noble, A. C. (1990). Bitterness and astringency in wine. In R. L. Rouseff (Ed.).Bitterness in foods and beverages (Vol. 25, pp. 145–158). Amsterdam: Elsevier.

Nurgel, C., & Pickering, G. (2006). Modeling of sweet, bitter and irritant sensationsand their interactions elicited by model ice wines. Journal of Sensory Studies,21(5), 505–519.

Oberholster, A., Francis, I. L., Iland, P. G., & Waters, E. J. (2009). Mouthfeel of whitewines made with and without pomace contact and added anthocyanins.Australian Journal of Grape and Wine Research, 15(1), 59–69.

Peleg, H., Gacon, K., Schlich, P., & Noble, A. C. (1999). Bitterness and astringency offlavan-3-ol monomers, dimers and trimers. Journal of the Science of Food andAgriculture, 79(8), 1123–1128.

Pérez-Magariño, S., Ortega-Heras, M., & Cano-Mozo, E. (2008). Optimization of asolid-phase extraction method using copolymer sorbents for isolation ofphenolic compounds in red wines and quantification by HPLC. Journal ofAgricultural and Food Chemistry, 56, 11560–11570.

Preys, S., Mazerolles, G., Courcoux, P., Samson, A., Fischer, U., Hanafi, M., et al.(2006). Relationship between polyphenolic composition and some sensoryproperties in red wines using multiway analyses. Analytica Chimica Acta, 563(1–2 SPEC ISS.), 126–136.

Ribereau-Gayon, P. (1970). Le dosage des composés phénoliques totaux dans lesvins rouges. Chemie Analytique, 52(6), 627–631.

Robichaud, J. L., & Noble, A. C. (1990). Astringency and bitterness of selectedphenolics in wine. Journal of the Science of Food and Agriculture, 52,343–353.

Sáenz-Navajas, M. P., Avizcuri, J. M., Ferreira, V., & Fernández-Zurbano, P. (2012).Insights on the chemical basis of the astringency of Spanish red wines. FoodChemistry, 134(3), 1484–1493.

Sáenz-Navajas, M. P., Campo, E., Avizcuri, J. M., Valentin, D., Fernández-Zurbano, P., & Ferreira, V. (2012). Contribution of non-volatile and aromafractions to in-mouth sensory properties of red wines: Winereconstitution strategies and sensory sorting task. Analytica Chimica Acta,732, 64–72.

Sáenz-Navajas, M. P., Campo, E., Fernández-Zurbano, P., Valentin, D., & Ferreira, V.(2010). An assessment of the effects of wine volatiles on the perception of tasteand astringency in wine. Food Chemistry, 121(4), 1139–1149.

Sáenz-Navajas, M. P., Ferreira, V., Dizy, M., & Fernández-Zurbano, P. (2010).Characterization of taste-active fractions in red wine combining HPLCfractionation, sensory analysis and ultra performance liquid chromatographycoupled with mass spectrometry detection. Analytica Chimica Acta, 673(2),151–159.

Sáenz-Navajas, M. P., Martín-López, C., Ferreira, V., & Fernández-Zurbano, P.(2011). Sensory properties of premium Spanish red wines and theirimplication in wine quality perception. Australian Journal of Grape and WineResearch, 17(1), 9–19.

198 A. Gonzalo-Diago et al. / Food Chemistry 154 (2014) 187–198

Sáenz-Navajas, M. P., Tao, Y. S., Dizy, M., Ferreira, V., & Fernández-Zurbano, P.(2010). Relationship between nonvolatile composition and sensory propertiesof premium Spanish red wines and their correlation to quality perception.Journal of Agricultural and Food Chemistry, 58(23), 12407–12416.

Scharbert, S., & Hofmann, T. (2005). Molecular definition of black tea taste by meansof quantitative studies, taste reconstitution, and omission experiments. Journalof Agricultural and Food Chemistry, 53(13), 5377–5384.

Scharbert, S., Holzmann, N., & Hofmann, T. (2004). Identification of the astringenttaste compounds in black tea infusions by combining instrumental analysis andhuman bioresponse. Journal of Agricultural and Food Chemistry, 52(11),3498–3508.

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, 1525–1533.

Stevens, J. C. (1997). Detection of very complex taste mixtures: Generousintegration across constituent compounds. Physiology and Behavior, 62(5),1137–1143.

Sun, B., Leandro, M. C., de Freitas, V., & Spranger, M. I. (2006). Fractionation of redwine polyphenols by solid-phase extraction and liquid chromatography. Journalof Chromatography A, 1128(1–2), 27–38.

Tepper, B. J., White, E. A., Koelliker, Y., Lanzara, C., D’Adamo, P., & Gasparini, P.(2009). Genetic variation in taste sensitivity to 6-n-propylthiouracil and its

relationship to taste perception and food selection. Annals of the New YorkAcademy of Sciences, 1170, 126–139.

Vidal, S., Courcoux, P., Francis, L., Kwiatkowski, M., Gawel, R., Williams, P., et al.(2004). Use of an experimental design approach for evaluation of key winecomponents on mouth-feel perception. Food Quality and Preference, 15(3),209–217.

Vidal, S., Francis, L., Guyot, S., Marnet, N., Kwiatkowski, M., Gawel, R., et al. (2003).The mouth-feel properties of grape and apple proanthocyanidins in a wine-likemedium. Journal of the Science of Food and Agriculture, 83(6), 564–573.

Vidal, S., Francis, L., Noble, A., Kwiatkowski, M., Cheynier, V., & Waters, E. (2004).Taste and mouth-feel properties of different types of tannin-likepolyphenolic compounds and anthocyanins in wine. Analytica Chimica Acta,513(1), 57–65.

Vidal, S., Francis, L., Williams, P., Kwiatkowskia, M., Gawel, R., Cheynier, V., et al.(2004). The mouth-feel properties of polysaccharides and anthocyanins in awine like medium. Food Chemistry, 85, 519–525.

Weber, F., Greve, K., Durner, D., Fischer, U., & Winterhalter, P. (2013). Sensoryand chemical characterization of phenolic polymers from red wine obtainedby gel permeation chromatography. American Journal of Enology andViticulture, 64, 1.

Whiting, G. C. (1976). Organic acid metabolism of yeasts during fermentation ofalcoholic beverages-a review. Journal of the Institute of Brewing, 82, 84–92.