Abstract
Background and goals Most premium wines are produced from grapes of Vitis vinifera parentage. Concord and related Vitis labruscana cultivars are typically considered unacceptable for premium wine production due to the musky aroma of their juice, often described as “foxy”. In this work, we compared the sensory attributes of a wine produced from nanofiltration-resin (NFR)-treated Concord juice to the sensory attributes of a standard Concord wine, a V. vinifera wine, and an NFR:V. vinifera blend. The volatile composition of NFR and Concord wines was evaluated to determine the extent to which the “foxy” character of Concord juice could be lessened with NFR, thus potentially improving the overall consumer liking of its wine.
Methods and key findings Wine was prepared from NFR-treated Concord juice and evaluated by a trained descriptive sensory panel and by untrained wine consumers alongside a standard Concord wine, a representative V. vinifera wine, and 75:25 blend of V. vinifera:NFR Concord wine. Both trained panelists and untrained consumers reported that the “foxy” character in the NFR Concord wine was significantly lower than that found in a standard Concord wine. Consumers reported an increase in overall liking of the NFR Concord wine over the standard Concord wine. Volatiles (n = 34) in the NFR Concord and standard Concord wine were quantified by gas chromatography-mass spectrometry. Most volatiles were similar between the two wines, with the largest difference observed for methyl anthranilate (“Concord grape” aroma), which was 96% lower in the NFR Concord wine.
Conclusions and significance Wine produced from NFR Concord juice had labrusca (“foxy”) character only slightly higher than V. vinifera wine. NFR pretreatment may expand winemaking options for Concord juice.
Introduction
Several odorants, including methyl anthranilate (MA), 2-aminoacetophenone (2-AAP), and furaneol, are responsible for the “labrusca” character of Vitis labruscana grape cultivars (Nelson et al. 1977), with descriptors often including attributes like “native”, “candy-like”, and/or “foxy”. In juices and wines from Concord grapes, the most widely planted V. labruscana cultivar in New York (as found on the website https://newyorkwines.org/grapes), the most important of these “foxy” odorants is MA. Concentrations of MA in Concord juice and wine are well above the compound’s detection threshold of 45 μg/L and 89 μg/L, in white wine and model wine, respectively (Perry and Hayes 2016). MA concentrations in Concord juice and wine may also be greater than the rejection threshold, reportedly 130 μg/L in white wine assessed by wine experts in California, and 1.7 mg/L in white wine assessed by nonexpert wine consumers in California (Perry et al. 2019).
Historically, Concord grapes have primarily been consumed as juice, with additional culinary uses in jams, jellies, and table wines. However, like many other fruit juices (e.g., apple), sales of grape juice have been in decline in recent years (Murray 2020) for several reasons, including consumers looking to limit their own or their children’s sugar consumption, and the increased availability of sugar-free or sugar-reduced beverage alternatives (Heyman et al. 2017).
In contrast to juice consumption, wine consumption has increased since the early 1990s (as found on the website https://wineinstitute.org/our-industry/statistics/us-wine-consumption). A potential alternative use for Concord grapes would be to increase their utilization in wine production beyond sweet bulk wine, especially because the price of Concord grapes is considerably less than European winegrapes (Vitis vinifera) in many regions. For example, the average cost of Concord in 2023 was $450/ton in the Finger Lakes region of New York, as compared to an average of $1900/ton for V. vinifera grapes (as reported in https://bpb-us-e1.wpmucdn.com/blogs.cornell.edu/dist/0/9318/files/2023/09/2023-Grape-Prices.pdf), with Concord vines being both higher yielding and having lower input costs (Davis et al. 2020).
Although Concord is currently used in wine production, increasing its utilization is made challenging by its labrusca character because of the presence of MA and “foxy” aromas. For certain regions and demographics, the presence of MA is of little concern—for example, nonexpert wine consumers in Pennsylvania did not reject wines with added MA (Perry et al. 2019), possibly because consumers were familiar with Concord flavor in local wines, juices, and jams. Similarly, Canziani et al. (2018) noted that consumers were much more likely to purchase Muscadine wine if they had a history with Muscadine-based products such as juice, jams, and jellies. However, for many consumers, MA in Concord is present at concentrations above its rejection threshold (Perry et al. 2019), and “foxy” and “native” are generally considered pejorative terms among wine professionals (Robinson 2015). For Muscadine, many consumers believe there is a quality stigma attached to such wines (Duarte Alonso 2011), and similar beliefs are expected regarding incorporation of labrusca-type grapes into premium wine production.
In contrast to many other wine odorants that are either formed de novo by yeast (e.g., esters, higher alcohols) or by reactions involving grape-derived precursors during fermentation and storage (e.g., isoprenoids, S-conjugate derived thiols), MA is present in grapes and does not appear to be precursor-derived (Waterhouse et al. 2024). Potentially, prefermentation removal of MA and related “foxy” odorants from Concord could render it appropriate for premium wine production. By analogy, Ryona et al. (2012) demonstrated that using a nonpolar silicone sorbent selectively decreased 3-isopropyl-2-methoxypyrazine (IPMP; “green pea” aroma) in grape juice. Fermenting the juice resulted in a wine with decreased concentrations of IPMP, but with negligible effects on most other wine volatiles. However, silicone is not an approved material for treatment of wine and juice in the United States and other countries, and the approach would be cumbersome on a larger scale.
Nanofiltration of Concord juice followed by resin treatment of the permeate (NFR) was proposed as an approach to remove MA without affecting other key components (e.g., color, sugars, and acids) (Ortiz Quezada et al. 2023). Although membrane filtration coupled with adsorption approaches have been reported for removal of unwanted components in wine, such as removal of Brettanomyces-produced volatile phenols (Ugarte et al. 2005), this strategy has not been extended to juice. In this work, we evaluated the hypothesis that wines produced from NFR-treated Concord juice should have reduced MA and “foxy” aromas comparable to a representative V. vinifera wine. We also evaluated the hypothesis that consumer liking of a wine from NFR-treated juice should be greater than liking of a standard Concord wine.
Materials and Methods
Chemicals
Standards of MA (≥98%), 2-octanol (≥97%), and β-damascone (≥95%) were purchased from Sigma Aldrich. Isotopically labeled internal standards, hexanal-d12 (≥98% d12), linalool-d3 (≥99% d3), and vanillin-d3 (≥99% d3) were purchased from CDN Isotopes; methyl anthranilated3 (MA-d3; ≥95% d3) was purchased from Eptes. Sodium chloride (NaCl), phosphoric acid, and sodium hydroxide (NaOH) were purchased from VWR; potassium metabisulfite (KMS) was purchased from L.D. Carlson; and hydrogen peroxide was purchased from a local grocer.
Model juice (16 Brix, 10 g/L tartaric acid, pH 3.5) and model wine (10% v/v ethanol, 10 g/L tartaric acid, pH 3.5) were made using deionized water, tartaric acid (Avantar), and either sucrose (Domino cane sugar) or 70% ethanol (Koptec), with pH adjusted using NaOH (5.0N, VWR).
Juice treatments and fermentations
A Concord juice concentrate (68 Brix) produced through evaporative processes without essence returned was sourced from Welch’s Inc. The concentrate was diluted with deionized water to 16 Brix to lower osmotic pressure and improve flux during nanofiltration, with total soluble solids (TSS) measured using a handheld refractometer (ATAGO PAL-1). The juice was processed as outlined by Ortiz Quezada et al. (2023). Briefly, Concord juice (4-L batches, n = 3 replicates) was loaded into a feed tank and circulated through a laboratory-scale crossflow reverse osmosis mini filtration system (Tangent Membranes) containing a nanofiltration spiral-wound membrane made of polyamide thin-film composite (MWCO 600-800 Da, Synder Filtration). The retentate was immediately returned to the feed tank. The permeate passed through a cartridge containing Amberlite XAD16N resin (Sigma Aldrich) to adsorb MA prior to its return to the feed tank. The system was operated at 4137 kPa at 30°C. Each of the three NFR processing runs treated juice for 3 hr; juices from the three runs were blended prior to fermentation. To produce a Concord wine with characteristic labrusca-type aromas for use as a standard, a commercial Concord juice (Welch’s 100% Grape Juice Concord Grape, 16 Brix) was purchased locally. Both juices were chaptalized to ~22 Brix to reach a desired alcohol level of between 12 and 13.5% ABV.
Juices were then fermented in triplicate (Figure 1). Yeast (0.3 g/L, ICV-D80, Lalvin) were rehydrated with Go-Ferm Protect Evolution (0.3 g/L, Cellar Science). Glass ferment jugs (3.8 L) were 75% filled with juice such that the fill level was just below the curvature of the top of the jug. After rehydration, yeast was pitched, and jugs were fitted with air locks. Fermaid O (0.2 g/L, Lallemand) was added in two stages: first, after the lag phase, and second, after 50% sugar depletion. Fermentation progress was monitored daily by recording TSS levels (in units of Brix) and temperature (°C), using a handheld density meter (DMA 35, Serial 82667610, Anton Paar). Ferments were allowed to proceed until yeast had flocculated and settled to the bottom of each jug and bubbling in the airlock had ceased. Sulfur dioxide (SO2) was added in the form of KMS to the finished ferments to achieve a molecular SO2 level of 0.6 mg/L. For clarification, wines were held in carboys at room temperature, then racked off the fine lees, with free SO2 levels monitored over this time using either automated titration against a standard iodine solution (HI 901W, ORP electrode, Hanna Instruments), or manually by aeration oxidation as described in Iland et al. (2013). Once wine was racked off the fine lees, experimental replicates were blended to ensure adequate quantities of wine for sensory analysis, then hand-bottled in 375-mL green glass bottles (Waterloo Container) under a stream of nitrogen (Airgas) and corked with Nomacorc synthetic #9 3.81 cm corks (Model No RK-4K08-ZN9Q, Home Brew Ohio) using an Italian floor corker (Ferrari). Bottles were stored upright at room temperature until analyses of basic wine chemistry, sensory, and volatiles were performed.
Summary of the experimental design, outlining the processing and vinification steps of concentrates and juices. The base concentrate used for nanofiltration-resin (NFR) processing was Concord concentrate without essence (i.e., Concord juice without volatiles added back after concentration via evaporative methods). The Concord (V. labruscana) wine was vinified from commercially available Concord juice; each of the juices were fermented in triplicate and then blended before bottling.
Juice and wine composition analysis
Juice samples were frozen in a commercial −20°C freezer and analyzed at the Cornell Craft Beverage Analytical Lab (Geneva, NY) for TSS (Brix) by refractometry, pH by pH meter, and titratable acidity (TA; g/L expressed as tartaric acid equivalents) by titrating to pH 8.2 using an autotitrator and pH meter. Yeast assimilable nitrogen was determined as the sum of the primary amino nitrogen and ammonium, measured by enzymatic analysis.
Wine samples were analyzed for basic chemistry at the Cornell Craft Beverage Analytical Lab. Ethanol was measured by gas chromatography-flame ionization detection, pH was measured with a pH meter, TA was measured using an autotitrator set to pH 8.2, and SO2 and residual sugar were measured by enzymatic assays. Wine samples were diluted two-fold in McIlvaine’s buffer (pH 3.2, LabChem), and absorbances at 420 and 520 nm were determined by an Agilent 8453 diode-array UV-visible spectrophotometer. Intensities were calculated by summing the absorbance values at 420 and 520 nm, while hue values were calculated by dividing the absorbance values at 420 nm by the absorbance values at 520 nm.
Sensory testing
All sensory testing was reviewed and approved by Cornell University’s Institutional Review Board for Human Participant Research (protocol #1510005908). All participants provided informed consent and were compensated for their time.
Sensory evaluations
Wine preparation
Four wines were used: i) a standard Concord wine fermented from commercial Concord juice (“V. labruscana”), ii) a wine fermented from NFR-processed Concord juice (“NFR”), iii) an unoaked commercial Gamay noir made using traditional vinification methods (“V. vinifera”), and iv) a 75:25 blend of the purchased Gamay noir wine and NFR-processed Concord (“blend”). The Gamay was purchased from a local liquor store and stored at room temperature until required for the test. The wine chosen was a commercial 2020 unoaked Beaujolais-Villages (100% Gamay noir) from Domaine de la Madone, which was produced using standard winemaking techniques and not carbonic maceration, according to the producer’s website. The wine “blend” was made before the start of the test by blending 75% of the purchased Gamay with 25% of the NFR wine. The blended wine was then put into 750-mL bottles to enable 30-mL pour spouts to be used to aliquot the wine. NFR and Concord wines were stored at room temperature under cork closure until the start of the test.
Trained panel
A panel of 11 Cornell graduate students, ranging from 24-to 30-years-old, was enlisted from a larger trained panel for descriptive analysis. Panelists practiced rating wines using a line scale from 0 to 10, with indented anchor points placed at 4, 7, and 10, where 4 was defined as “slight”, 7 defined as “definite”, and 10 defined as “pronounced”. Sensory training was administered over six weeks, comprising a minimum of 17 total sessions on top of the panel’s existing descriptive sensory training, with each session lasting between 30 and 60 min. During each session, panelists were trained to recognize aroma standards (listed in Supplemental Table 1) by using three concentrations of the aroma standard used to train panelists on aroma intensity. Taste and mouthfeel training standards were also administered during the sessions (listed in Supplemental Table 2), including sweet, bitter, sour, and tannic. Concentrations and reference standards were selected based on previous work (Reyes et al. 2017, Ko et al. 2020).
Once panelists had completed basic sensory training, in which each of the aromas were presented in isolation in a water matrix over the course of two sessions, the listed aromas were added to a lab-fermented neutral Chardonnay wine; panelists underwent further training to identify the aromas in the wine matrix for remaining training sessions. Throughout the training period, three tests were administered to panelists to gauge the progress and performance of each panelist. The first test required naming the correct attribute in a single-attribute spiked wine, the second test required correct identifications of a double-attribute spiked wine, and the third test required correct identification of all three attributes spiked into a single wine. Performance on the tests guided focus and compound revisitation in subsequent sessions.
Specific to this study, the aroma standard of MA was given to panelists over nine training sessions to train panelists on the attribute of “foxy”. Initially, the concentrations used to train panelists were 50, 150, and 500 μg/L for the “slight”, “definite”, and “pronounced” anchors, respectively. However, after the first two sessions, the panel unanimously decided that the concentrations needed to be lowered and subsequent sessions used adjusted concentrations of 40, 100, and 250 μg/L as “slight”, “definite”, and “pronounced”, respectively.
A comprehensive descriptive analysis (DA) was conducted in duplicate on all four wines, over two sessions administered in subsequent days, to characterize attribute intensities of the wines. Because of pandemic health guidelines, the test was administered as an “at-home” study. Sensory samples were prepared the morning of distribution to panelists. High-purity nitrogen gas was used to purge 60-mL glass bottles. Each wine (30 mL) was poured into 60-mL glass bottles. High-purity nitrogen was then used to purge the headspace of the bottle before a plastic cap was screwed onto the wine-filled bottle. A blinding code sticker was placed on each bottle. Wines for session one and session two were placed in separate bags and labeled as “Day 1” and “Day 2”, respectively. Blinding codes for the wines were different for Day 1 and Day 2. Panelists were instructed to complete the test within 72 hr of receiving their kit. Data were collected using RedJade sensory evaluation software, which could be accessed online. All panelists were familiar, and had previous experience, with at-home sensory tests. Panelists were not provided with wine glasses but were instructed to use four identical glasses for the study.
On each of the two days for descriptive evaluation, panelists were instructed to pour the wine into a clean and empty glass, smell the sample, and rate the intensity of each of the following attributes: floral, yeasty, citrus, foxy, fruity, green bell pepper, and oxidized/vinegar. Panelists rated intensities using a sliding horizontal line scale with anchors at 0 (none), 10 (barely detectable), 40 (slight), 70 (definite), and 100 (pronounced). Panelists were then instructed to taste the wine and evaluate the intensity of sweetness, bitterness, tannic, and sour. Panelists used the same line scale with the same anchors provided for the aroma attributes to rate the taste and mouthfeel intensities. Panelists were asked to cleanse their palate with tap water and unsalted crackers between samples, with a forced break time of 10 sec.
Consumer acceptance testing
Consumer acceptance testing occurred one month after DA. A total of 95 consumers were recruited through Cornell’s Sensory Evaluation Center (SEC). Inclusion criteria required participants to be of legal drinking age (at least 21-years-old), compliant with COVID-19 protocols, and report liking red wine. Consumers attended a single tasting session at the SEC, equipped with eight individual booths under fluorescent lighting. During the session, consumers tasted 30 mL of each of the four wines: i) the standard Concord, ii) NFR-processed Concord, iii) unoaked commercial Gamay, and iv) 75:25 blend of commercial Gamay and NFR-processed Concord, presented in a counterbalanced order. Wines were served in ISO-standard tasting wine glasses covered with plastic lids and labeled with three-digit blinding codes. Wines were served at room temperature (20 to 23°C). Consumers received one wine at a time and were instructed to score the color of the wine, where 0 = light and 100 = dark. Consumers were then asked to sample the wine and indicate their overall liking, as well as their liking of the wine’s appearance, aroma, flavor, mouthfeel, and aftertaste, using a nine-point scale where 1 = dislike extremely, 5 = neither like nor dislike, and 9 = like extremely. Consumers were also asked to rate their willingness-to-pay (WTP) and to rate the intensity of “foxiness” of the wine using a vertical sliding scale from 0 to 100, where “foxiness” was described as “the aroma from Concord/Niagara grapes, similar to the aroma of Welch’s grape juice/jelly, grape cough syrup, and grape candies”. Lastly, consumers were asked a series of questions to capture their demographics. Data were collected using RedJade software installed on each of the computers in the SEC booths.
Semiquantitative volatile analysis by solid phase microextraction-arrow (SPME-arrow) coupled with gas chromatography-mass spectrometry (GC-MS)
Samples were prepared for analysis by GC-MS by transferring 10 mL of sample into a 20-mL amber vial containing 3 g of NaCl. Samples were diluted 100-fold for the analysis of ethyl isovalerate, isoamyl acetate, ethyl hexanoate, 2-methyl-1-butanol, isoamyl alcohol, ethyl octanoate, linalool, phenylethyl alcohol, and octanoic acid. For the analysis of the remaining 25 volatiles, the samples were diluted ten-fold. Internal standards were added to each sample such that the concentration of the internal standards in solution were 200 μg/L MA-d3, 200 μg/L hexanal-d12, 50 μg/L 2-octanol, 50 μg/L linalool-d3, 25 μg/L β-damascone, and 25 μg/L vanillin-d3. Samples were promptly analyzed by GCMS following preparation.
Samples were loaded onto the autosampler (LEAP CombiPAL) and volatiles were extracted by headspace SPME-arrow (10 mm, 80 μm divinylbenzene-carbon wide range-polydimethylsiloxane DVB/CWR/PDMS; RESTEK, batch 220060) using a preextraction incubation temperature of 60°C for 1 min with 500 rpm agitation followed by a 5 min extraction. The SPME-arrow was desorbed for 5 min in a split/splitless injector of a Shimadzu gas chromatograph-mass spectrometer (GC2010 Plus w/ TQ8040MS) in splitless mode at a constant temperature of 280°C or 230°C (for juice or wine, respectively), a sampling time of 2 min, and a purge flow of 3 mL/min.
Juice samples were analyzed with a RESTEK Rxi-5Sil MS column (30 m × 0.25 mm × 0.25 μm) with helium as a carrier gas at a flow rate of 1.4 mL/min. The GC oven was held for 2 min at 35°C, then followed the sequence: 7°C/min ramp to 42°C, 11°C/min ramp to 75°C, 7°C/min ramp to 170°C, 120°C/min ramp to 280°C, then a 3 min hold at 280°C. The total GC program run time was 23.49 min. The MS was operated in electron ionization (EI) mode with an ionization energy of 70 eV and a solvent cut time of 3.75 min. MS data were collected over the range m/z 45 to 250 in full scan mode with an effective sampling rate of 3 Hz.
Wine samples were analyzed with a RESTEK Stabilwax-MS column (30 m × 0.25 mm × 0.25 μm) with helium as a carrier gas at a flow rate of 1.4 mL/min. The GC oven was held for 2 min at 35°C, then followed the sequence: 7°C/min ramp to 42°C, 10°C/min ramp to 75°C, 7°C/min ramp to 170°C, 120°C/min ramp to 250°C, then a 3 min hold at 250°C. The total GC program run time was 23.54 min. The MS was operated in EI mode with an ionization energy of 70 eV and a solvent cut time of 5 min. MS data were collected over the range m/z 45 to 250 in full scan mode with an event time of 0.1 sec, in addition to single-ion mode for m/z 151 (MA) and 154 (MA-d3) with an effective sampling rate of 33 Hz.
Postrun data processing was performed using Shimadzu GCMS Solution ver. 4.50 software. MA was identified based on the retention properties of a pure MA standard. Juice calibration standards of MA (10, 100, 500, 1000 μg/L) in model juice and wine calibration standards of MA (10, 50, 100, 1000 μg/L) in model wine were prepared, with MA-d3 (200 μg/L) as an internal standard, to generate calibration curves for the quantification of MA based on the ratio of peak areas for MA (m/z 151) and MA-d3 (m/z 154). The calibration curves were generated using weighting factors of ([concentration]−1 or 1/[concentration value]).
Beyond MA and its deuterated standard, an additional 33 compounds were quantified in GC-MS runs based on their high signal intensity and library match quality. Tentative identifications were based on spectral similarity and Kovats’ retention indices using the NIST online database. Quantification for each compound was determined using the major ion for each compound. If the major ion was saturated or there were interferences with the ion, the ion with the next highest signal was used for quantification. The concentration of each volatile was calculated using the peak areas of each of the compounds of interest. For semiquantitation, the relative response was defined as the peak area for each compound of interest (abbreviated CI) compared to the peak areas of each of the major quantifying ions of the five internal standards (Equation 1); vanillin-d3 was omitted because of poor identification using the method described in this work.
The relative standard deviation (RSD), Equation 2, was calculated for each internal standard for every CI to determine which internal standard would minimize the RSD for each CI.
Once the internal standard for each CI was determined, the ratios of (CI : corresponding internal standard) were compared between Concord wine and NFR wine.
Statistical analyses
JMP Pro ver. 16.0.0 (SAS Institute Inc.) was used for all statistical analyses, with α = 0.05. Briefly, two-sided t-tests were performed to evaluate significance in the basic chemistry data (e.g., ethanol, acidity) and for determining significance of the color data.
Linear mixed models were used to analyze sensory data sets from both the panel and consumers. The two sessions by the trained panelists were accounted for in the model by setting replicates as a random effect for the panel data. For the consumer test, because consumers rated more than one product and therefore each observation was not independent of one another, a linear mixed model was chosen over a one-way analysis of variance to determine differences in overall liking between the four wines, where “product” was a fixed effect and “consumer” was a random effect. Linear correlation statistics within the data set were calculated as Pearson’s r (p < 0.05). Linear mixed models were also chosen to analyze possible effects of self-reported wine training experience and wine consumption on overall liking.
Differences in volatile concentrations in the standard Concord and NFR Concord wines were evaluated using individual t-tests. Statistically significant differences were based on a Bonferroni corrected p value of < 0.0015 (0.05 significance level/34 tests performed).
Results
A standard commercial Concord juice and an NFR-treated Concord juice were fermented in triplicate; all juices fermented to apparent dryness (<0 Brix). The Concord ferments reached apparent dryness after 7.6 ± 0.1 day and the NFR after 12.0 ± 2.4 day. Because of the different sources of the juices, the slower fermentation kinetics of the NFR juices could not be ascribed solely to the NFR treatment.
The initial chemistry of the NFR juice and the comparison Concord juice showed negligible differences; the resulting wines were similar in basic chemistry as well (Supplemental Table 3). Both Concord and NFR had similar final alcohol (~13.5% v/v). Both wines reached dryness, with no significant differences between residual sugar in NFR (1.0 ± 0.3 g/L) and Concord wines (0.5 ± 0.06 g/L) (0.05 < p < 0.1). The finished Concord wines had significantly lower acid (pH 3.39 ± 0.010, TA 6.01 ± 0.06 g/L) than the NFR-processed wines (pH 3.26 ± 0.006, TA 7.04 ± 0.06). Because the commercial Concord and NFR juices were from different sources, these differences in basic wine composition could not be assigned solely to the NFR treatment. Both wines showed a marginal increase in TA compared to their juice counterparts, which could be due to formation of succinic acid and other acids produced by yeast (Fowles 1992). NFR and Concord wines differed in color, with the NFR having higher abs 420, abs 520, and color intensity, along with lower hue (Supplemental Table 4).
Trained panel sensory evaluations
A trained panel evaluated four wines for sensory characteristics: the Concord wine, the NFR-treatment wine, a commercial unoaked red V. vinifera wine (100% Gamay), and a 75:25 blend (commercial Gamay wine:NFR-treatment wine). The “foxiness” intensity for each of the four wines is shown (Figure 2A). Panelists rated the Gamay and blend wines the lowest in perceived “foxiness” (medians of 1.05 and 9.65, respectively), the NFR wine higher (median of 39.7), and the Concord wine the highest (median of 69) (p < 0.05). Linear mixed models for the data indicate that there were no significant effects (p > 0.05) due to panelist or session (Wald p values of 0.076 and 0.667, respectively).
“Foxiness” intensity in each of the four wines (scale 0 to 100) shown for A) trained panelists (n = 11, duplicate evaluations) and B) consumers (n = 95). The four wines were an unoaked commercial Gamay noir made using traditional vinification methods (“V. vinifera”); a 75:25 blend of the purchased Gamay noir wine and nanofiltration-resin (NFR)-processed Concord (“Blend”); a wine fermented from the NFR-processed juice (“NFR”); and a standard Concord wine fermented from the commercial Concord juice (“V. labruscana”). The histograms show the number of responses, with the box plots illustrating the interquartile range of the "foxiness" ratings. Outliers of "foxiness" ratings are represented with a bullet (•) and the mean for each wine is marked by ×. Letters indicate significance using a linear mixed model, incorporating “wine type” as a fixed effect, and “panelist” and “session” as random effects (p < 0.05).
In addition to “foxy” aroma, the trained panel also evaluated the wines for six other aroma attributes and four taste and mouthfeel attributes (Figure 3) on a sliding 100-point scale. Panelists did not find differences for most attributes, including floral, yeasty, citrus, and oxidized aromas. There were also no perceived differences in astringency or sourness, despite small differences in pH and TA (Supplemental Table 3). Significant differences were found for fruitiness, bell pepper, sweetness, and bitterness (p < 0.05). The NFR wine was perceived as fruitier than the blended wine (54 versus 37). Green bell pepper aroma also differed; the Gamay wine had a higher intensity of green bell pepper aroma than the NFR and Concord wines, although scores for this trait were low for all wines (15 for the Gamay wine versus 5 for the NFR and Concord wines). NFR wine was perceived as sweeter than the blended wine and Gamay wine (35 versus 18 and 21, respectively), while NFR was perceived as less bitter than the blended wine and Gamay wine (24 versus 37 and 39, respectively) (p < 0.05).
![Intensity of seven aroma attributes (floral, yeasty, citrus, "foxy", fruity, green bell pepper [GBP], and oxidized/vinegar) and four taste and mouthfeel attributes (sweetness, bitterness, tannic, and sour) of four wines evaluated by trained panelists (n = 11, duplicate evaluations). The four wines were an unoaked commercial Gamay noir made using traditional vinification methods (“V. vinifera”); a 75:25 blend of the purchased Gamay noir wine and nanofiltration-resin (NFR)-processed Concord (“Blend”); a wine fermented from the NFR-processed juice (“NFR”); and a standard Concord wine fermented from the commercial Concord juice (“V. labruscana”). An asterisk (*) indicates significant differences (p < 0.05) among wines within an attribute, with different letters indicating which wines were significantly different from each other for that given attribute.](https://www.ajevonline.org/content/ajev/76/1/0760002/F3.large.jpg?width=800&height=600&carousel=1)
Intensity of seven aroma attributes (floral, yeasty, citrus, "foxy", fruity, green bell pepper [GBP], and oxidized/vinegar) and four taste and mouthfeel attributes (sweetness, bitterness, tannic, and sour) of four wines evaluated by trained panelists (n = 11, duplicate evaluations). The four wines were an unoaked commercial Gamay noir made using traditional vinification methods (“V. vinifera”); a 75:25 blend of the purchased Gamay noir wine and nanofiltration-resin (NFR)-processed Concord (“Blend”); a wine fermented from the NFR-processed juice (“NFR”); and a standard Concord wine fermented from the commercial Concord juice (“V. labruscana”). An asterisk (*) indicates significant differences (p < 0.05) among wines within an attribute, with different letters indicating which wines were significantly different from each other for that given attribute.
Consumer sensory evaluations
Overall liking of the four wines used in the descriptive study was evaluated by a consumer panel (n = 95) using a nine-point hedonic scale (Figure 4). The standard Concord wine received the lowest overall liking score (4.86), while the overall liking ratings for the other three wines (5.90 to 6.19) were significantly higher than the standard Concord wine (p < 0.05) but did not differ significantly from each other. Overall liking scores for the Concord wine had a bimodal distribution with maxima at “3” and “7”, but liking for the other wines appeared unimodal (Figure 4). The NFR sample was perceived as less “foxy” than the Concord wine (median of 20.7 versus 52.3) but was still more “foxy” than the V. vinifera (Gamay) (median 13.9), with the blend (median 16.9) indistinguishable from either the Gamay or NFR wines (Figure 2B).
Overall liking by consumers (n = 95, scale = 1 to 9) of four wines: an unoaked commercial Gamay noir made using traditional vinification methods (“V. vinifera”); a 75:25 blend of the purchased Gamay noir wine and nanofiltration-resin (NFR)-processed Concord (“Blend”); a wine fermented from the NFR-processed juice (“NFR”); and a standard Concord wine fermented from the commercial Concord juice (“V. labruscana”). The histograms show the number of responses, with the box plots illustrating the interquartile range of the overall liking ratings. Outliers of liking ratings are represented with a bullet (•) and the mean for each wine is marked by the ×. Letters indicate significance (p < 0.05), with Concord wine rated lower in overall liking compared to the other three wines.
In addition to overall liking and “foxiness”, several other subattributes related to liking—color, aroma liking, flavor liking, mouthfeel liking, aftertaste liking, appearance liking, and WTP—were also assessed. To determine if any terms were used independently of each other, regressions and correlation coefficients were determined for all pairwise comparisons (Supplemental Figure 1). All terms were covariant with overall liking (r > |0.178|, p < 0.05); “foxiness” was the only term inversely correlated with overall liking (r = −0.212, p < 0.05). All liking terms were covariant with WTP (r ≥ 0.237, p < 0.05); however, color and “foxiness” did not covary with WTP (r < |0.05|, p > 0.05).
Volatile analysis by GC-MS
Concentrations of MA in the lab-fermented wines (x̄ ± sx) were 878 ± 67 μg/L in the standard Concord (V. labruscana) and 36 ± 6 μg/L in the NFR-treated Concord, compared to <20 μg/L measured in the commercial Gamay (V. vinifera). In addition to MA, several other volatiles (n = 34) were tentatively identified in Concord and NFR wines and semiquantified by GC-MS: seven alcohols, 19 esters, four fatty acids, one aldehyde, two monoterpenes, and one C13norisoprenoid. Relative changes of MA and these volatiles in Concord and NFR wine are shown (Figure 5). In addition to MA (−96% in NFR wine), there were a few volatiles that showed sizable (>80%) differences: hexanol (−96% in NFR wine as compared to Concord wine), hexyl acetate (−98%), methyl octanoate (−86%), and ethyl nonanoate (−81%). Only one compound, decanol (+86%), was significantly higher in the NFR-treatment wine. Of the volatiles shown in Figure 5, 44% (15 of 34) showed differences of <40%, including several branched and straight-chain ethyl esters (e.g., ethyl hexanoate, ethyl isovalerate), fatty acids (e.g., hexanoic acid), and higher alcohols (e.g., isoamyl alcohol).
Differences in the volatile profile of wines fermented from Concord juice versus nanofiltration-resin (NFR)-processed Concord. Volatiles were measured semiquantitatively by gas chromatography-mass spectrometry. The scale shows the relative decrease of volatiles in the NFR wine compared to the standard Concord wine; only decanol, ethyl-2-methylbutyrate, and ethyl isovalerate were found in higher concentrations in the NFR wine compared to the Concord wine. Significance is indicated with an asterisk (*), with a Bonferroni corrected p value < 0.0015. TDN, 1,1,6-trimethyl-1,2-dihydronaphthalene.
Discussion
Sensory analysis
In sensory experiments, we compared the NFR wines against two other wines, a representative “foxy” wine produced from commercial Concord juice using the same protocols as the NFR Concord, and a representative “non-foxy” commercial Gamay V. vinifera wine. A 75:25 blend of the commercial Gamay wine, NFR-treatment wine was also included in sensory evaluations, as such a blend could be labeled as a varietal wine in the United States (27 CFR §4.23). The sensory data collected from both trained panelists and wine consumers indicate that “foxiness” was much higher in the standard Concord wine than in the other three wines (NFR treatment, commercial V. vinifera, blend) (p < 0.05). Previous studies have shown that the sensory threshold for MA varies depending on the matrix (Perry and Hayes 2016), as well as on the consumer’s familiarity with wine and Concord-based products (Perry et al. 2019). However, the range of sensory thresholds for MA reported in the previous studies (45 to 130 μg/L) is comparable to the MA concentration in the NFR wine, and well below the concentration measured in the standard Concord. Thus, the lower “foxiness” of the NFR wine is likely due to removal of MA, although removal of other odorants associated with “foxiness” cannot be ruled out, e.g., o-aminoacetophenone is also reported to contribute “foxiness”, but was not measured in this study because of its low concentration. Notably, a 75:25 blend of the Gamay and NFR wines had “foxy” character comparable to the 100% V. vinifera wine (Figure 2), suggesting that NFR-treated Concord may be useful as an inexpensive blending component at levels allowed for varietal wines without imparting “foxy” aromas.
Other sensory terms evaluated by the trained panel did not differ significantly among wines, or else showed only minor differences. Notably, panelists did not perceive the NFR-treatment wine as any more or less fruity than the V. labruscana (standard Concord) wine (p > 0.05). This observation is described in more detail in the next subsection.
The sensory data collected from consumers suggests that overall, consumers preferred the treated wine with a lower concentration of MA (overall liking was less for the Concord wine, p < 0.05) (Figure 4). Predictably, overall liking correlated with WTP (Supplemental Figure 1). However, although overall liking correlated with perceived “foxiness”, the “foxy” rating did not correlate with WTP. As noted above, the distribution of liking for the Concord wine in the current study was bimodal. Previous work has shown that consumers’ interest in wine (“Low” versus “High”) is inversely correlated with their preference for “foxy” flavors, such that Low Interest consumers have a higher rejection threshold for MA (Perry et al. 2019). The current study lacked sufficient information to segment consumers based on Low versus High Interest. However, we can speculate that the bimodal WTP and liking data in the current study were further evidence of this segmentation of consumers based on wine interest, with the caveat that there were other differences between the NFR and V. vinifera wines beyond “foxy” character.
Volatile composition
GC-MS analyses confirmed that the NFR Concord wine had much lower MA than the standard Concord wine, to levels below its sensory threshold in wine. MA in wines is reported to exist in grapes only in its free form, and no grape-derived precursors have been identified (Waterhouse et al. 2024). Thus, the strong correlation between initial juice MA and final wine MA was unsurprising.
Several other volatiles were also detected in NFR Concord wines. Most of these volatiles are commonly reported fermentation metabolites (especially acetate esters, straight chain and branched chain ethyl esters, fatty acids, and higher alcohols). Although the NFR Concord juice and standard Concord were derived from different starting juices, they were fermented under identical conditions, and we observed negligible or nonsignificant differences among most fermentation volatiles. For example, higher alcohols (e.g., 2-methylbutanol, isoamyl alcohol, phenylethanol) are largely formed from yeast amino acid metabolism, especially via anabolic pathways (Nisbet et al. 2014, Rollero et al. 2017), and minimal differences in these compounds were found across treatments (<20%) (Figure 5). Similarly, straight-chain saturated fatty acids (e.g., hexanoic acid and octanoic acid) and their corresponding ethyl esters (ethyl hexanoate and ethyl octanoate, respectively) are by-products of yeast fatty acid synthesis, and are produced almost entirely from yeast metabolism, with little contribution from the grape (other than as a source of fermentable sugars) (Nisbet et al. 2014). Because these volatiles are produced de novo after the NFR treatment, it is unsurprising that the standard and NFR Concord wines have similar concentrations of these compounds.
Other than MA, a few other volatiles (ethyl 2-butenoate, ethyl 2-hexenoate) detected in wines in this work are present in Concord grapes (Schreier and Paroschy 1981, Ortiz Quezada et al. 2023). Monoterpenes (linalool, α-terpineol) and 1,1,6-trimethyl-1,2-dihydronaphthalene (TDN; a C13norisoprenoid) were also detected in wines, and these are expected to be produced mostly from glycosidase activity and acid-catalyzed rearrangements of glycoside or polyol precursors during fermentation and/or storage (Ferreira and Cacho 2009). Modestly lower concentrations were observed for these compounds (<40%) in the NFR wine. One possible explanation is that precursors of these compounds were partially removed by the NFR treatment, although this cannot be confirmed because of the different origins of the juices. Similarly, the lower concentration of hexanol in the NFR wine is consistent with the findings from Ryona et al. (2012) using sorbent-treated juice, and also with 13C-labeling studies, which showed that most hexanol in wine is not produced by fermentation of sugars (Nisbet et al. 2014), but rather from enzymatic oxidation of grape lipids (Waterhouse et al. 2024). Thus, hexanol could be lower in NFR wines because it is removed by the juice filtration treatment. Similarly, hexyl acetate was higher in the Concord wine than in the NFR wine. Hexyl acetate is formed from the reaction of grape-derived hexanol with acetyl-coenzyme A during fermentation (Styger et al. 2011), and its lower concentration in NFR Concord wine could be related to its lower hexanol concentration. As with the monoterpenes and TDN, this hypothesis cannot be confirmed because the starting juices had different sources.
In comparing the sensory and volatile data, the lower intensity of “foxiness” in NFR Concord than in standard Concord aligns well with the lower MA concentration of the former. However, the lower concentration of a few odorants associated with fruity aromas (e.g., hexyl acetate, phenethyl acetate) did not translate to lower perceived fruitiness in the NFR Concord. This is potentially because of masking effects of MA. Alternatively, the lack of effect on fruitiness may be because most fruit-smelling esters were unaffected or showed negligible changes. De la Fuente-Blanco et al. (2020) concluded that 14 ethyl esters common in wine aroma can be replaced by three, or even one, critical ethyl ester. Additionally, many of the volatiles important for the “aroma buffer” (Ferreira et al. 2022) that impart the characteristic “vinous” aroma to wine (e.g., isoamyl alcohol, phenylethyl alcohol, hexanoic acid, octanoic acid, ethyl octanoate, ethyl decanoate) were mostly unaffected.
Limitations and scope
This work compared the sensory attributes of the NFR Concord wine to a standard Concord wine and a commercial Gamay. Although both trained and consumer panels reported lower “foxy” attributes in the NFR Concord than in the standard Concord, these differences cannot be definitively assigned to the NFR treatment because of the different origins of the juices. However, the large reduction in MA observed in NFR Concord does suggest that this is a plausible cause of the sensory differences. Additionally, the Gamay was selected because it was a fruity wine without oak influence, and thus appropriate as a V. vinifera benchmark for this experiment, but it is only one of many traditional European winemaking grapes. Different sensory results for some attributes (e.g., floral, yeasty, citrus) would be observed if an alternative wine was chosen as the representative V. vinifera, although we would likewise expect low “foxiness” in other V. vinifera wines.
The differences in appearance among the wines may have also had an effect on the sensory results. The COVID-19 pandemic necessitated both at-home testing for the DA and the use of standard clear, colorless wine glasses. There were significant differences between color measurements (abs 420, abs 520, intensity, hue) for all four wines (Supplemental Table 4), with the Gamay wine having the highest color intensity. Although it is possible that color influenced the consumer or trained panel responses, such data is useful in understanding how a product like the NFR wine would be perceived in a traditional tasting room setting.
Conclusion
Trained and consumer sensory panels reported lower “foxiness” in the wine produced from Concord juice processed by NFR treatment as compared to a standard Concord (V. labruscana) wine, which may be due to removal of MA by the NFR treatment. The consumer sensory panel preferred the NFR Concord wine, suggesting that NFR treatment of Concord juice may yield Concord wine more suitable for use in premium wine production.
Supplemental Data
The following supplemental materials are available for this article in the Supplemental tab above:
Supplemental Table 1 Panelists were trained on the aroma standards listed in this table, first in water and subsequently in an unoaked Chardonnay wine. An asterisk (*) indicates that listed concentrations were modified from initial concentrations after panel feedback. † indicates that the concentration was calculated according to acetic acid concentration in apple cider vinegar (~5%).
Supplemental Table 2 Taste and mouthfeel standards, prepared in water and used to train sensory panelists, adapted from Ko et al. (2020) and Reyes et al. (2017). An asterisk (*) indicates that the concentrations provided had been lowered from higher concentrations after feedback from the panelists.
Supplemental Table 3 Basic chemistries for the Concord (“V. labruscana”) and nanofiltration-resin (NFR)-processed juices and wines (“NFR”, “Blend”, and “V. vinifera”). Fermentation replicates (n = 3) were measured, with the standard deviation of the samples provided. TSS, total soluble solids; TA, titratable acidity (in g/L tartaric acid equivalents); PAN, primary amino nitrogen; AMM, ammonia; YAN, yeast assimilable nitrogen; RS, residual sugar. Statistical significance within each column is indicated with different letters, based on a student’s t-test (p < 0.05).
Supplemental Table 4 Color values for the final wines were measured using UV-vis spectroscopy, recording absorbance values at 420 and 520 nm. Samples were diluted two-fold with McIlvaine’s buffer (pH = 3.2), with a path length of 1 cm. Values provided are measurements of the diluted samples, with means listed ± 1 standard deviation (n = 3). Statistical significance within each column is indicated with different letters, based on a one-way analysis of variance followed by a Tukey’s honest significant difference post-hoc test (p < 0.05). NFR, nanofiltration-resin processing.
Supplemental Figure 1 Scatterplots of the sensory data from consumers (n = 95) for four wines (an unoaked commercial Gamay noir made using traditional vinification methods, a 75:25 blend of the purchased Gamay noir wine and nanofiltration-resin (NFR)-processed Concord, a wine fermented from the NFR-processed juice, and a standard Concord wine fermented from the commercial Concord juice) evaluated for color intensity (0 to 100), overall liking (1 to 10), appearance liking (1 to 10), aroma liking (1 to 10), flavor liking (1 to 10), mouthfeel liking (1 to 10), aftertaste liking (1 to 10), “foxiness” (1 to 100), and willingness to pay (WTP). The points represent individual observations; points have been made moderately transparent to allow data density to show through for the discrete variables. The density ellipses outline a 95% bivariate normal density. The degree of correlation between variables is reflected in the narrowness/roundness of the ellipse (narrow and diagonal ellipse correspond to a greater degree of correlation). The labels on the significance circles are the correlation values, corresponding to the color of the circle. The size of the circle indicates correlation probability; the size is inversely proportional to the negative log10 of the p value (i.e., a larger circle indicates a smaller p value). An asterisk (*) indicates a correlation probability < 0.05.
Footnotes
Funding was provided by New York State Department of Agriculture and Markets (Award #C00441GG and CM04068H), New York Wine and Grape Foundation (Enology I-9), and Welch Foods Inc. We acknowledge the assistance of Heather Scott in reviewing the manuscript and Glycine Jiang in training the sensory panel.
Perry DM, Ortiz Quezada AG, Guan W, Dando R and Sacks GL. 2025. Using sensory evaluation and volatile analysis to determine the enological potential of Concord juice processed by nanofiltration-resin. Am J Enol Vitic 76:0760002. DOI: 10.5344/ajev.2024.24046
By downloading and/or receiving this article, you agree to the Disclaimer of Warranties and Liability. If you do not agree to the Disclaimers, do not download and/or accept this article.
The data underlying this study are available on request from the corresponding author.
- Received August 2024.
- Accepted November 2024.
- Published online February 2025
This is an open access article distributed under the CC BY 4.0 license.
References
Vol 76 Issue 1
Jump to section
Related Articles
Cited By...
More from this TOC section
Similar Articles
