Standards of 20 L-amino acids, as well as of acetonitrile and formic acid, were purchased from Sigma-Aldrich. All standards had a purity level that exceeded 99.5 %, and the solvents were LC/MS grade. Ammonium formate was obtained from Fluka. Milli-Q water was produced using a water purification system from Merck.

In 2021 and 2022 (the two study vintages), Chenin blanc grapes were harvested at a maturity level selected by the winemaker collaborating on the project; the grapes came from three different vineyards (hereafter called A, B and C) with surface areas of about 1 ha located in the Loire Valley of France. The soil in vineyard A was mainly sandstone schist with sandy loam on the surface; the soil in vineyard B was mainly schist with sandy loam on the surface; and the soil in vineyard C was mainly volcanic (spilite) with sandy clay on the surface.

The 2021 vintage saw early bud break due to a mild winter, followed by spring frosts and cool, wet conditions in summer, slowing down the growth and ripening of grapes. In contrast, the 2022 vintage, marked by spring frosts but a hot, dry summer, favoured early and uniform ripening, with low disease pressure despite water stress partially offset by summer rains.

The barrels used in the study were Pure T (low toasting) as well as 172M and 179M (medium toasting); they measured 400 or 500 L and were obtained from Taransaud ^®. Neither barrel production type nor volume was considered in this study.

For each vintage, grapes were harvested at maturity (at around 22 °Brix and a pH of 3.2). In 2021, the °Brix, pH, and titratable acidity (eq. H ₂SO ₄ g/L) values were 23.0, 3.3, and 5.10 respectively for vineyard A; 23.0, 3.2, and 5.62 for vineyard B; and 21.6, 3.1, and 6.84 for vineyard C. In 2022, they were 21.2, 3.4, and 4.17 for vineyard A; 23.9, 3.3, and 4.53 for vineyard B; and 22.1, 3.3, and 4.26 for vineyard C.

Before the study, the winemaker had selected yeasts from his vineyards and then prepared inocula using these selected yeasts and autochthonous yeasts. These inocula were added to different musts: the 02/10, 27/09, and 29/09 musts in 2021 for vineyards A, B and C, respectively, and the 17/09 musts in 2022 for vineyard A and the 11/09 musts in 2022 for vineyards B and C. After settling and at the start of alcoholic fermentation (AF), the musts were transferred into new barrels (NB) and old barrels (OB), in which fermentation and ageing proceeded. The wine cellar was kept at a temperature of 20 °C during fermentation (AF and malolactic fermentation (MLF)). Following AF, density values were measured. MLF was spontaneous and was monitored using periodic analyses of malic and lactic acid levels. AF and MLF reached completion about a month after inoculation for AF, marking the start of ageing. Ten and eleven wines (representing the combinations of vineyards × barrels) for 2021 and 2022, respectively, were analysed over the 29 weeks of ageing. The study started after AF and MLF had been completed ( i.e., week 6 for the 2021 vintage and week 7 for the 2022 vintage) and finished 29 weeks later ( i.e., week 35 for the 2021 vintage and week 36 for the 2022 vintage). This study applied a partial factorial design, because it was conducted under real-life conditions in an actual winery with low yields (Figure 1): old barrels were used in all combinations (since old barrels are the most common ones used in the region), and new barrels were used in a subset of combinations.

Une image contenant texte, capture d’écran, diagramme, affichage Le contenu généré par l’IA peut être incorrect.

In 2021, 15 mg/L of OptiFlore ^® (Lamothe–Abiet) was added to the must of vineyard B, whose level of assimilable nitrogen was only 142 mg/L. In 2022, the musts of all three vineyards had assimilable nitrogen levels of around 100 mg/L. They were thus supplemented with NUTRISTART ^® (Laffort): with 40 g/hL in the case of vineyards B and C and with 50 g/hL in the case of vineyard A. Just after MLF had finished, 1 g/hL of SO ₂ was added to all the wines. During ageing, light stirring was performed approximately four times. All the wines were collected in 75 cL bottles, after very light stirring to avoid moving the lees in the deepest part of the barrels. Upon arrival in the laboratory, samples of the wines were distributed among 50 mL Eppendorf tubes and stored at –70 °C.

Amino acid analyses were carried out using an Agilent LC 1260 Infinity II System with a mass detector and a simple quadrupole Agilent LC-MSD iQ System. Chromatographic separation was performed using an Agilent InfinityLab Poroshell 120 HILIC-Z (150 × 2.1 mm, 2.7 µm particle size) with a guard column. The two mobile phases were each composed of 10 % ammonium formate buffer (pH 3.4) and either 90 % water (mobile phase A) or 90 % acetonitrile (mobile phase B). A gradient mixture of the two mobile phases was used in the analysis (flow rate of 0.4 mL/min at 30 °C) as follows: 0–10.5 min: 0 to 11 % of A; 10.5–15 min: increase to 12 % of A; 15–24 min: increase to 25 % of A. Three microlitres were injected into the chromatographic system.

The amino acids present (Table 1) were detected by mass spectroscopy. Electrospray ionisation (spray voltage: 3,500 V) was applied, and SIM spectra were recorded in positive ion mode. During the analysis, the source temperature was 350 °C, the fragmentation voltage was 80 V, and the gas flow was fixed at 8 L/min. A stock solution (the 20 amino acid standards in the A mobile phase) was used for calibration. Before the injection, each solution was filtered through a PTFE 0.22 µm filter. Calibration was carried out on the mass spectra using the specific ion extracted chromatograms and employing nine points of calibration.

The method has been validated by the fact that all amino acids were clearly identified thanks to their RT and their m/ z. The calibration curves were all linear ( R ² ≥ 0.99 except for Arg, 0.97) except for Pro, a quadratic equation was preferred ( R ² = 1.00). The LOD (Limit of Detection) and LOQ (Limit of Quantification) are available in Table S1. All standards were injected at days 1, 2, and 3 and no significant difference was observed. The variation coefficients of the repeatability of the standard preparation and injection comprised between 0 and 4 %, and 6 % was reached for Asp with respect to the repeatability of the preparations.

The wines were thawed the day prior to running the analyses and were then kept at 4 °C. The analyses were carried out on three 50 mL samples of each wine that had been centrifuged and filtered (PTFE 0.22 µm filter) beforehand.

The replicates comprised the wines in the new barrels and the wines in the old barrels that had undergone the various combinations of vintage and vineyard in the partial factorial design.

Two-way ANOVAs were performed for each vintage-vineyard combination where the concentration of a given amino acid was the response variable, and the explanatory variables were sampling period ( i.e., week of ageing) and barrel age. When sampling period was significant, post-hoc pairwise comparisons (Bonferroni correction) were conducted to identify which periods differed significantly. All the tests employed a 95 % confidence level.

To evaluate all the variables simultaneously, a series of linear mixed-effects models was performed, where the concentration of a given amino acid was the response variable; vintage, vineyard, and sampling period (and their interactions) were the fixed effects; and barrel age was a random effect. Barrel age was treated as a random effect, because repeated measurements were obtained for the same barrels over time and barrel number was limited. This approach aimed to account for the unbalanced design resulting from the fact that not all factor combinations were present, which prevents the inclusion of all higher-order interactions in a standard ANOVA. All data analysis was performed using R software.

There were clear differences in amino acid concentrations between the two vintages (Table 2), all other factors combined (sampling period, barrel age, and vineyard).

First, total amino acid concentrations were 962.29 ± 237.27 mg/L in 2021 and 569.77 ± 93.87 mg/L in 2022. These values fell within the expected ranges of 235–1,812 mg/L and 417–2,121 mg/L, respectively, according to Mirás-Avalos et al. (2020) and Moreno-Arribas et al. (1998).

Asp occurred at very high concentrations in the wines of both vintages (435.30 mg/L in 2021). It was also the amino acid with the highest relative levels, since it accounted for 45 % and 59 % of total amino acid content in 2021 and 2022, respectively. Pro was the second most abundant amino acid in terms of concentration (299.23 mg/L in 2021) and accounted for 31 % and 20 % of total amino acid content in 2021 and 2022, respectively. The third most abundant amino acid was Glu (34.04 mg/L in 2021), which accounted for 4 % and 5 % of total amino acid content in 2021 and 2022, respectively. These three amino acids ( i.e., Asp, Pro, and Glu) accounted for 80–83 % of the total amino acid content in wines of both vintages.

All the amino acids had significantly higher concentrations in the 2021 wines than in the 2022 wines, indicating the effect of vintage (Table 3; p < 0.05); the only exceptions were Asp and Arg (Table 2).

Previous studies have observed an overall effect of vintage on amino acid concentrations (Ferrari & Feuillat, 1988), which appears to be particularly related to weather conditions (Ortega‐Heras et al., 2014). Local climate conditions in combination with soil composition (notably available nitrogen levels) can induce different grape ripening levels, as has already been shown for red grapes (Hernandez-Orte et al., 1999; Garde-Cerdán et al., 2018). To compensate for that effect, grapes were harvested later in 2021 than in 2022. The ripeness level was generally higher in 2022 than in 2021, suggesting that grape maturity could not explain the differences in amino acid concentrations observed in this study.

It is important to note the relatively large standard deviation values (Table 2), in particular those for Arg (40–48 %), Gln (34–36 %), His (38–44 %), and Lys (31–37 %) for both vintages and those for Pro (40 %) and Ile (45 %) for the 2021 vintage. This fact suggests heterogeneity in the data that is linked to the different combinations of factors, including vineyard, sampling period, and barrel age.

Patterns were seen in the amino acid concentrations in the wines from 2021 (Tables 4, S2 and S3) and the wines from 2022 (Tables 5, S4 and S5) The significance ( p-value) of the factors (sampling period, barrel age, and their interactions) are indicated in the tables as well.

First, wines of the two vintages and the three vineyards aged in old barrels were considered.

Some amino acids displayed the same patterns across vintages and vineyards. Ser concentrations increased significantly across the 29 weeks of ageing (+46 %). For instance, this rise was 61 % in 2021 wines from vineyard C. While Glu, Lys, Phe, and Thr concentrations displayed different patterns over the course of ageing, they were all significantly higher at the end than at the beginning across the two vintages and three vineyards. For example, Glu concentrations rose by 40 % in 2021 wines from vineyard A (Table 4). Thr is a precursor of sotolon, a lactone associated with nutty, curry, and oxidised aromas (Thuy et al., 1995). Although Thr concentrations generally increased over the 29 weeks of ageing, there were periods of decline, probably due to acid degradation; the latter could potentially have led to the formation of volatile compounds like sotolon in the presence of acetaldehyde, after a step of deamination (Thuy et al., 1995). Acetaldehyde could have been produced by the oxidation of ethanol in the presence of H ₂O ₂, itself generated by O ₂ oxidation in the presence of copper or iron, and in the absence of ascorbic acid (Barril et al., 2016). The absence of ascorbic acid was indeed verified. In red wines, acetaldehyde react with phenolic compounds (Han et al., 2019). In a white wine matrix with a low content of phenolic compounds, acetaldehyde might accumulate longer.

Certain amino acids displayed inconsistent patterns over the course of ageing. Gly concentrations generally increased, except in 2022 wines from vineyards A and C, for which there was no change. Similarly, Tyr concentrations generally increased, except in 2022 wines from vineyards A and B, in which there was no change; His and Asp concentrations generally increased, except in 2022 wines from vineyard B and in 2021 wines from vineyard A, in which they were unchanged. Fiechter and Mayer (2011) also observed an increase in Asp concentrations among control wines and wines that had aged three and six months on lees and that had been made using three different fermentation yeasts and Grüner Veltliner grapes; similar results were seen by Ferrari and Feuillat (1988) in Chardonnay. Val concentrations generally increased, except in 2022 wines from vineyard A and in 2021 wines from vineyard B, in which there was no change. Val is one of the main amino acids released by peptide hydrolysis (Alexandre et al., 2001). Asn concentrations generally increased, except in 2022 wines from vineyard A, in which a decrease was seen, and in 2022 wines from vineyard C, in which no change was seen.

Consequently, total amino acid content ( i.e., the sum of all the amino acid concentrations) also significantly increased during ageing: from 12.7 % (2021 vineyard B) to 35.4 % (2022 vineyard B), the greatest increases being in 2022 wines. The release of nitrogen indicates that there was autolytic activity linked to protease activity during ageing (Kemp et al., 2015). Indeed, in model solutions, Asp, His, Thr, and Ser have been seen to appear during the autolysis of lees (Alexandre et al., 2001). The occurrence of Thr and Ser could reflect the degradation of mannoproteins that had been released by glucanase activity during ageing (Alexandre et al., 2001). In addition, Asn, Asp, Gln, Glu, Ser, and His were found to originate from the activity of peptide hydrolyses (Alexandre et al., 2001) and from passive diffusion through the cellular membrane. This could explain the variability of the changes in Asn, Gln, and His concentrations over the course of ageing.

Two amino acids generally decreased in concentration over the course of ageing: Ala and Trp. For Ala, the decrease occurred in all wines, with the exception of 2022 wines from vineyard B. Previous research on sparkling wine has found Ala concentrations to slightly decrease between three and six months of ageing, followed by steady values (Sun et al., 2024). Trp concentrations also significantly decreased in all wines, with the exception of 2021 wines from vineyards A and B.

For other amino acids, concentration patterns were consistent within vintage across vineyards but varied between vintages. Arg concentrations significantly increased over ageing in 2022 wines from all vineyards. In 2021 wines, the pattern was more varied: Arg concentrations decreased between week 6 and week 28 and then increased between week 28 and week 36, with final concentrations that were equal to concentrations in week 6 in the case of vineyard B or that were lower in the cases of vineyards A and C. In 2021 wines, Pro concentrations significantly increased over the course of ageing; in 2022 wines, they dropped between week 7 and week 24 and then remained relatively steady.

Met concentrations increased over the course of ageing in 2021 wines; meanwhile, in 2022 wines, they were stable in the cases of vineyards A and B, but decreased significantly in the case of vineyard C. Met is the precursor of methionol ( i.e., 3-methylthio-1-propanol), which imparts a boiled or cooked potato or cauliflower aroma in wine, and of methanethiol, which has sewage and rubber aromas (Cordente et al., 2021). The former is generated via the Ehrlich pathway while the latter derives from a demethiolation step. Competition occurs between the enzymatic and the non-enzymatic reactions. The authors showed the impacts of the strains used for AF, methionol and methanethiol during Chardonnay ageing: the concentration of methionol was stable between 3 and 15 months of aging, whereas the concentration of methanethiol increased (Cordente et al., 2021). The consumption of Met generated by the Strecker degradation has been shown in wine conditions in the presence of gallic acid, caffeic acid, or catechin (Oliveira et al., 2017). That reaction needs a first step consisting in the oxidation of phenolic compounds into quinones. In addition, Met could also have been adsorbed by cell walls of lees via its -SH group by creating bisulfite bridge with the Cys of mannoproteins as thiols do (Fornairon-Bonnefond et al., 2001). The decreased concentrations in 2022 wines from vineyard C could have resulted from Met’s conversion into volatile compounds.

A clear vineyard effect, independent of vintage, was seen in Ile concentrations ( p = 1.85.10 ^–8): they were stable in wines from vineyard A, decreased in wines from vineyard B, and increased in wines from vineyard C. From a study concerning model wine solutions close to our wines, the decrease of Ile could be due to non-enzymatic reaction leading to 2-methylbutanal having fruity odours (Pripis-Nicolau et al., 2000). Aldehydes could be formed by Strecker degradation, which is a slow reaction.

Gln concentrations were relatively stable in wines from vineyards A and B. They significantly increased in 2021 wines from vineyard C up until week 13; they then decreased, but still remained higher than they had been at week 5. In 2022 wines, Gln concentrations declined during ageing. The patterns of Gln concentrations in 2021 wines from vineyard C were relatively consistent with results seen in Niagara and Goethe sparkling wines from the 2015 vintage (Sartor et al., 2021): Gln concentrations increased in Chardonnay between three and six months of ageing and displayed a strong decrease afterwards; i.e., after the second fermentation had finished. It may be possible to explain this heterogeneity by the fact that Gln is mainly the product of peptide hydrolysis (Alexandre et al., 2001), and wine protein and peptide contents are linked to vintage ( i.e., weather and vine health) rather than grape ripeness ( i.e., ripening level indicator values).

Finally, it was not possible to explain the concentration patterns of one amino acid by vintage effect or vineyard effect alone: Leu concentrations generally increased over the course of ageing in 2021 wines from vineyards B and C, but remained steady in 2021 wines from vineyard A and in 2022 wines from all vineyards. Indeed, the statistical analysis showed that there was a significant interaction between all the factors (Table 3). The decrease in Leu concentrations during ageing could be due to the amino acid’s conversion into volatile compounds (Sun et al., 2024), and in particular into 3-methylbutanal – which is associated with an amylic odour – by non-enzymatic reactions, as suggested by the work of Pripis-Nicolau et al. (2000).

The effect of barrel age is complex, since amino acid concentrations were influenced by vintage, vineyard, and sampling period, and their interactions (Table 3), making it difficult to explain the resulting patterns. However, the dynamic shifts in amino acid concentrations seen for the two vintages could be explained by a previously described phenomenon (Fornairon-Bonnefond et al., 2001; Guilloux-Bénatier & Chassagne, 2003; Gnoinski et al., 2021) that comprises different stages. First, yeast starvation (due to the lack of reducing sugars) induces autophagy, which generates macromolecule degradation within cell vacuoles and an increase in membrane porosity. Autophagy seems to activate autolysis. Second, compounds in the cytoplasm, including peptides, fatty acids, nucleotides, and amino acids, are released through the cell membrane by passive diffusion. Third, after a lag period, cell wall constituents, such as proteins and polysaccharides, are released by the irreversible action of proteases, with glucanases releasing mannoproteins. Fourth, peptides are hydrolysed, generating new free amino acids in the medium (Alexandre et al., 2001). Alexandre et al. (2001) proposed that the release of amino acids starts during the stationary growth phase of AF and is associated with cell death (Kemp et al., 2015). Enzymatic activities are slow, which implies that ageing durations ranging from months to years could have a significant effect on wine quality (Gnoinski et al., 2021). Tudela et al. (2012) observed visible changes in the ultrastructure of sparkling wine ageing on lees after 18 months, but the cell walls remained unbroken even after 40 months. Fornairon-Bonnefond et al. (2001) reported different studies that indicate that enzymatic activities still occur during ageing; for example, proteases have been found to be active until five to seven months, but they showed maximum activity after two months (Alexandre et al., 2001); meanwhile, the activity of esterases has been found to be significant only after 20 days on lies, and β-glucanases has shown activities for up to five months (Gnoinski et al., 2021). This suggests that enzymatic reactions may occur during ageing, but probably to release amino acids, peptides and polysaccharides rather than to directly produce volatiles. Nonetheless, the consumption of amino acids seems to be mainly due to non-enzymatic reactions that result in the production of volatiles, from Ile, Leu, and Met (Pripis-Nicolau et al., 2000). Further explanation cannot be given due to the lack of data concerning the evolution of amino acids during still-wine ageing on lees. The results of this study suggest that the decrease in concentrations of some amino acids over the course of ageing could be linked to their conversion to other molecules, like volatile compounds, or to their adsorption to the lees before being released again, due to further membrane enlargement.

Similar trends were observed in amino acid concentrations over the course of ageing, whether wines were in new or old barrels (vineyard A: Table 4 for 2021 and Table 5 for 2022; 2022 vineyard B: Table S4; 2022 vineyard C: Table S5).

Examining the results in greater detail, it is clear that barrel age did not significantly affect Phe and Trp concentrations during ageing or at the end of ageing, regardless of vineyard or vintage. However, Gly concentrations were significantly higher in wines aged in old barrels versus new barrels, regardless of vineyard or vintage; this difference was mainly observed at the end of ageing.

Barrel age had a significant effect on the concentrations of the main amino acids, regardless of vintage or vineyard, but its impact was rarely independent of sampling period (sampling period-by-barrel interaction; Tables 3, 4, 5, S4, and S5). Barrel age appeared to have the greatest influence on His, Trp, Ile, and Thr. Their later concentrations were not correlated with their initial concentrations at the start of ageing.

The results are unclear and they suggest that wine amino acid composition does not depend on whether the barrel is new or old, since it is likely that interactions occur between initial wine composition, period of ageing, barrel type, and barrel age. The main difference between new and old barrels is the oxygen transfer rate, which is higher for new barrels. Lees consume oxygen (Salmon et al., 2000; Pons-Mercadé et al., 2021), but oxygen is not considered to have an impact on autolysis at its start, given that protein degradation occurs, which protects wine from oxidation (Pons-Mercadé et al., 2021). Oxygen spurs the production of free radicals via chemical reactions, which generate metabolites that are cytotoxic (Gnoinski et al., 2021), leading to lower levels of cell viability and the earlier occurrence of autolysis (Salmon et al., 2000).

Once ageing had ended (at week 35 or 36), one of two main results were obtained: 1) barrel age (NB/OB) had no difference on the evolution of amino acid concentrations; 2) wines in old barrels had globally higher amino acid concentrations than did wines in new barrels (except for four values). Therefore, the hypothesis that there was more oxygen in the new barrels may be an explanation for the latter results. The presence of oxygen in contact with lies generates oxygen peroxide (Salmon et al., 2000), which can react with phenolic compounds leading to quinones (Barril et al., 2016). These quinones can react with amino acids to produce volatile compounds (Oliveira et al., 2017). Since the levels of phenolic compounds in a white wine and the concentration of oxygen are relatively limited, the aforementioned phenomenon may have occurred, but with a moderate magnitude, hence the changes in low amino acid concentrations. Another explanation could be that new and old barrels affected wine composition during the two stages of fermentation, leading to differences in wine composition at the start of ageing that were maintained over the course of aging. While this may explain most of the results when the sampling period and barrel age interaction was not significant ( p > 0.05; e.g., Gln, Glu, and His concentrations in 2022 wines from vineyard C), it does not explain the other results, notably Asp concentrations in 2022 wines from vineyard B. In the latter case, wines in the old barrels exhibited higher Asp concentrations in week 7 than did wines in new barrels, but this relationship was reversed in week 36, showing the impacts of ageing.