The aromatic component of wine is an important element in consumer wine appreciation, as olfactory preferences are generally driven by the intensity and the dominant family of aroma perceived at tasting (Lesschaeve, 2008). In the 1980s, Noble et al. (1987) undertook to classify wine aroma into families leading to the development of a “wine aroma wheel”, which became a reference for wine writers, consumers, producers and researchers. From this work twelve distinct aroma families were identified, including a “vegetative” family related to sensory attributes, such as “grass cut green”, “eucalyptus”, “mint” and “bell pepper”.

The molecular determinism of these “vegetative” notes in wine has been widely studied and is currently well-understood. 3-Isobutyl-2-methoxypyrazine (IBMP) and 3-isopropyl-2-methoxypyrazine (IPMP) are two potent grape-derived molecules contributing to the bell pepper and asparagus character respectively (Bayonove et al., 1975). 3-alkyl-2-methoxypyrazines have been among the most studied aroma compounds since their first discovery in Cabernet Sauvignon (Bayonove et al., 1975), and some years later in Sauvignon (Augustyn et al., 1982). 1,8-cineole (also called eucalyptol) and its isomer 1,4-cineole, which impart eucalyptus and hay/dried herbs aromas respectively, have received a lot of attention from researchers in the past few years (Antalick et al., 2015; Capone et al., 2012; Poitou et al., 2017). Although these monoterpenes can have an endogenous origin, it is also possible for them to be introduced to grapes and wines from neighbouring plants (Capone et al., 2012; Poitou et al., 2017). C6-compounds (i.e., trans-2-hexenal or hexanal), which are formed in the early winemaking stages through enzymatic oxidation of lipidic precursors, are responsible for the smell of cut grass in wines (Cordonnier and Bayonove, 1981). During alcoholic fermentation, a part of these compounds is reduced by yeast to an alcohol form or, for trans-2-hexenal, is involved in the synthesis of 3-mercaptohexanol (Roland et al., 2010), which generally limits their sensory contribution to finished wines. Another monoterpene known as piperitone was recently identified as the main contributor of the mint nuances perceived in aged Bordeaux wines (Picard et al., 2016; Pons et al., 2016). It was also highlighted that p-menthane lactones can enhance the freshness and the mint odours perceived in these wines (Picard et al., 2017).

However, few research studies have been conducted to investigate consumer appreciation of the molecules responsible for these “vegetative” characters. The most significant study demonstrated that 1,8-cineole became undesirable in red wines above a concentration of 27.5 μg/L, thus encouraging winemakers to produce wines with moderate levels of this compound (Saliba et al., 2009). Without performing a formal consumer study, piperitone was considered a positive aroma compound, as the molecule was positively correlated with aging-bouquet typicality scores (Picard et al., 2016). 3-alkyl-2-methoxypyrazines, and notably IBMP, are usually seen as highly undesirable compounds by wine professionals and researchers. However, while the appreciation of methoxypyrazines has not yet been studied in red wines, moderate levels of IBMP can have a positive contribution to the aroma of Sauvignon wines (King et al., 2011; Rajchl et al., 2009).

The aim of this study was to formally assess the appreciation of IBMP in white and red wines by applying the consumer rejection threshold methodology (CRT) (Prescott et al., 2005). This valuable approach, based on the application of the paired preference test, allows the concentration of a molecule that gives a significantly lower preference to be determined. It has been previously applied to several compounds or flavours, including those that cause wine faults, such as 2,4,6-trichloroanisole (Prescott et al., 2005), 1,1,6-trimethyl-dehydronaphthalene also known as TDN (Ross et al., 2014), ethyl phenyl-acetate and phenyl acetic acid (Campo et al., 2012), 1,8-cineole (Saliba et al., 2009) and rotundone (Geffroy et al., 2018).

Sauvignon and Fer are two grape cultivars grown in the southwest of France within the Protected Designation of Origin (PDO) Gaillac, which cover 500 and 1000 ha respectively, according to the local winegrower association (www.vins-gaillac.com). Wines made from these two grape cultivars, which are known to be regular IBMP producers (Allen et al., 1991; Davaux, 2005; Geffroy et al. 2019; Lacey et al., 1991; Poitou et al., 2017), were used as base wines for this study. In both base wines, the best estimate threshold (BET) for olfactive detection of IBMP was also determined to find out the concentration at which the addition of IBMP can be perceived by panelists.

The two commercial unwooded base wines made from pure Sauvignon and Fer (PDO Gaillac) were selected on the basis of absence of any vegetative aroma at tasting. Such wines are typical of those produced in this area during the five last vintages, which are characterised by warm and dry climate conditions.

While performing the olfactory examination, the Sauvignon wine dominantly expressed floral aroma, while the Fer wine was characterised by black fruit and spicy notes. The determination of IBMP concentrations by Nyséos (Montpellier France), with SPME-GC-MS using deuterated IBMP and previously described parameters, (King et al., 2011) revealed negligible levels of this aroma compound: below the quantification limit of 1 ng/L in Sauvignon, and 3 ng/L in Fer. Classical analyses were also performed on the wines using a Winescan FT-120 (Foss France SAS, Nanterre, France), apart from free and total sulfur dioxide, which were determined according to an iodometric method using a Mi455 titrator (Milwaukee Instruments, Rocky Mount, USA). As reflected by the data summarised in Table 1, and notably the level of ethanol, the wines were produced from fully ripe grapes. This advanced stage of maturity, together with the fact that the two wines were sourced from vintages under warm climate conditions (2018 for Fer, 2019 for Sauvignon), must have contributed to limiting the level of IBMP in finished wines (Roujou de Boubée, et al. 2000).

Table

titre du tableau
Parameter	Sauvignon	Fer
Ethanol (% v/v)	14.6	14.9
Titratable acidity (g/L tartaric acid)	5.9	5.5
pH	3.29	3.62
Tartaric acid (g/L)	2.0	1.9
Malic acid (g/L)	0.6	0.1
Volatile acidity (g/L acetic acid)	0.62	0.50
Glucose/fructose (g/L)	0.4	1.7
CO2 (mg/L)	678	215
DO420	0.07	–
MCIa	–	12.5
TPIb	7.2	75.0
Free SO2 (mg/L)	22	11
Total SO2 (mg/L)	72	48
IBMP (ng/L)	<1	3

–not relevant for this product; aModified colour intensity; bTotal phenolic index

Food grade 3-isobutyl-2-methoxypyrazine (purity > 99 %) provided by Sigma-Aldrich (Saint-Quentin-Fallavier, France) was used to spike the wine for the BET and CRT studies. The spiked wines and the unspiked control wines were prepared 24 h before the tastings. The results of the analyses confirmed that the concentration of added IBMP was appropriate, with a recovery rate of 87–94 %. For all the tasting sessions, samples were coded with three-digit codes, and a constant volume of 15 mL was poured at 18 °C in black and clear wine-tasting glasses for the BET and CRT studies respectively. The tasting sessions took place in a neutral room with white walls and the spacing between the panelists ensured that no communication could occur. The BET and CRT studies were conducted in Toulouse, in the southwest of France, during the last week of February 2020 and the first week of March 2020 respectively.

The BET concentrations for olfactive detection of IBMP in Sauvignon and Fer wines were determined according to the American Society of Testing and Materials method E 679 (ASTM, 2004). The panelists (n = 20 for Sauvignon and n = 19 for Fer), were students from École d'Ingénieurs de Purpan aged between 22 and 25 years old, who had previous experience with difference testing and with sensory evaluation of wine. The percentage of female panelists was 65 % and 58 % for Sauvignon and Fer respectively. Due to the different number of male and female panelists, the ASTM E 679 method, which provides individual threshold values in a single session, was preferred to ASTM E 1432 to investigate the impact of gender on the ability to perceive IBMP. The concentrations to be tested were decided following preliminary tasting by a small group of panelists and on the basis of odour thresholds previously reported in the literature (Allen et al., 1991; Roujou de Boubée et al., 2000). The Sauvignon wine, containing undetectable level of IBMP, was spiked with 1, 2, 4, 8, 16 and 32 ng IBMP/L. The Fer wine, initially containing 3 ng IBMP/L, was spiked with 2, 4, 8, 16, 32 and 64 ng IBMP/L, corresponding to final concentrations of 5, 7, 11, 19, 35, and 67 ng/L. The 3-alternative forced choice (3-AFC) tests, with control wines incorporated, were served in ascending order of IBMP concentrations, as described in the E 679 method. Assessments were only conducted orthonasally (sniffing) without tasting. Panelists had to make a choice, even if no difference was noted. In accordance with the official method (ASTM, 2004), it was assumed that a panelist, who missed at the highest concentration, would have been correct at twice this level. Conversely, it was assumed that a panelist, who succeeded at the lowest concentration, would have been incorrect at half this level. The BET concentrations were first assessed for the Sauvignon wine, and after a 10 min rest for the Fer wine.

The panel was composed of 48 wine consumers recruited during a wine fair that took place at the Ecole d’Ingénieurs de Purpan in Toulouse, in the southwest of France. The participants were not offered any pay for participating in the study and they did not receive any information about the samples they had to evaluate. At the end of the session, they were asked to provide some demographic information and to respond to several questions regarding their purchase behaviour and their knowledge of vegetative notes in wine. The panel comprised 22 women and 26 men who were between 18 and 24 (31 %), 25 and 34 (23 %), 35 and 44 (4 %), 45 and 54 (23 %) and over 55 (19 %) years old. They reported their habits to be as follows: wine consumption for a mean of 16.4 years (standard deviation = 12.5), every day (6 %), 3 to 4 times a week (40 %), once a week (29 %), twice a month (15 %), and once a month (10 %). They described their level of knowledge as ‘no wine knowledge’ (23 %), ‘interested in wine’ (62 %), and ‘passionate about wine’ (15 %). The questionnaire showed that 56 % of the panelists had previously heard of vegetative aroma in wine, and that 46 % of them appreciated this kind of note.

The procedure for CRT determination was based on a replicate series of five paired comparison tests, one for each IBMP concentration (Prescott et al., 2005). Each pair consisted of a sample of the base wine alone and a sample of the base wine spiked with IBMP. The tasting order was randomised, and each pair was presented in order of ascending concentration. For each base wine, the concentrations were chosen based on the BET study, with the lowest concentration tested being close to the BET value. Panelists were asked to conduct an olfactory assessment, place the whole sample in their mouth, move it around for a few seconds and then spit it out. They then had to indicate the coded sample of the pair they preferred on a form. Panelists received a new pair of samples every five minutes.

Table

titre du tableau
	IBMP concentration (ng/L)	Gendera	Ageb	Level of wine knowledgeb	Familiarity with vegetative notes in winea	Self-reported liking of vegetative notes in wineb
Sauvignon	7.5	0.917	0.206	0.051	0.726	0.976
	15	1.00	0.811	0.357	0.251	0.050
	30	1.00	0.417	0.830	1.00	0.497
	60	1.00	0.728	0.227	1.00	0.563
	120	0.950	0.038	0.697	1.00	0.171
Fer	18	0.437	0.811	0.892	1.00	0.110
	33	0.281	0.732	0.694	0.839	0.078
	63	1.00	0.364	0.574	1.00	0.164
	123	1.00	0.956	0.389	1.00	0.750
	243	1.00	0.082	0.564	1.00	0.024

^aThe Kolmogorov-Smirnoff (two-tailed) test was performed; ^bThe Kruskall-Wallis test was performed; n = 48 consumers

For each panelist, the BET concentration was calculated as the geometric mean of the concentration at which the last miss occurred and the next higher concentration. The panel threshold was the geometric mean of the BET of the individual panelists, as described in the E 679 method (ASTM, 2004). A Mann-Whitney test was performed to detect any differences in BET according to gender.

The CRT concentration was calculated as the concentration at which a spiked wine was significantly rejected, using the binomial distribution for a paired comparison test (Roessler et al., 1978). The Kolmogorov-Smirnoff (two-tailed) test was used to evaluate whether the preference for IBMP differed according to gender and familiarity with vegetative notes in wine. For age, level of wine knowledge and self-reported liking of vegetative notes in wine, the Kruskall-Wallis test was performed. Man-Whitney, Kruskall-Wallis, and Kolmogorov-Smirnov tests were conducted with XLSTAT software (Addinsoft, Paris, France).

The panel BET was estimated as being 5.5 ng/L and 16.8 ng/L for Sauvignon and Fer respectively. These results are in accordance with previous studies, emphasising an odour threshold for IBMP of 8 ng/L in Sauvignon (Allen et al., 1991), and 15 ng/L in Cabernet Sauvignon, Cabernet Franc and Merlot (Roujou de Boubée et al., 2000). The importance of the matrix (i.e., the base wine) on aroma perception in wine has already been highlighted (Robinson et al., 2009; Villamor et al., 2013). These latter reported that increased contents in ethanol, tannin and sugar could reduce the headspace concentration of volatile aroma compounds and threshold values by between 2- and 10,000-fold. As the two studied matrices contained similar levels of ethanol and sugar (Table 1), we can suppose that the larger amount of tannins found in the Fer wine, as reflected by TPI values, is mainly responsible for the BET differences observed between the two wines. The differences in aroma composition between the two wines might also have played a role via mechanisms related to perception interactions (McKay et al., 2020).

As the panel was only composed of young panelists (< 25 years old), we cannot exclude the fact that slightly higher BETs would have been observed for a more age-balanced population. Indeed, it is well known that sensitivity for detecting odour usually declines with age (Griep et al., 1995). In the same way, as experience is known to facilitate the ability of panelists to discriminate wines when conducting triangular tests (Solomon, 1990), the fact that a less trained panel would have led to a higher BET cannot be rejected.

The distribution of best estimate threshold concentration showed a large variability between panelists and did not follow a normal law (Figure 1). For Sauvignon, one panelist succeeded at the lowest concentration and three panelists missed at the highest concentration. According to the Mann-Whitney test, no significant differences in BET was observed at P < 0.05 between female and male panelists for Sauvignon (P = 0.067) and Fer (P = 0.302). However, BETs tended to be lower for female panelists (3.5 ng/L for Sauvignon and 14.6 ng/L for Fer) in comparison to male panelists (21.6 ng/L for Sauvignon and 21.6 ng/L for Fer). Actually, most of the research investigating sex differences in terms of olfactory threshold have reported either greater female sensitivity or no sex differences in sensitivity, depending on the odorant (Doty and Cameron, 2009). Consequently, the fact that the greater proportion of female panelists within the panel might have contributed to lower the BETs for both wines cannot be completely discarded.

Table