2.1. Evolution of Phenolic Compounds
The profile of non-anthocyanin polyphenols is reported in
Table 1.
Ten non-anthocyanin phenolic compounds were identified. Although each of the four wines had a different phenolic profile due to the different composition of the grape varieties used for winemaking, a common evolution trend was detected for six out of 10 compounds, namely, gallic acid, caffeic acid,
p-coumaric acid,
trans-resveratrol, glutathionyl caftaric acid (GRP) and protocatechuic acid. During the first three months of bottle storage (T1, T2, T3) their concentrations did not show notable variations. At six months (T4), a net increase was observed, followed by a clear decrease at 12 months (T5). This trend was detected in the three red wines; however, in Lagrein rosé, this evolution was registered only for gallic acid, but not for the other phenolic compounds. A decrease of the low molecular weight compounds such as caffeic acid, (+)-catechin, (−)-epicatechin and
p-coumaric was observed when comparing the final concentrations (T5) with the initial ones (T1), as already reported in previous studies [
19,
24].
The anthocyanins were derivatives of delphinidin, cyanidin, petunidin, peonidin and malvidin, which are the typical anthocyanins present in wines obtained from
Vitis vinifera L. red grape varieties. Anthocyanins were grouped in three families according to the esterification groups: glucosides, acetyl-glucosides and coumaroyl-glucosides. Their concentrations are reported in
Table 2.
The glucosides-anthocyanins included delphinidin-3-
O-glucoside, cyanidin-3-
O-glucoside, petunidin-3-
O-glucoside, peonidin-3-
O-glucoside and malvidin-3-
O-glucoside. Acetyl-glucosides included delphinidin-3-
O-acetylglucoside, petunidin-3-
O-acetylglucoside, peonidin-3-
O-acetylglucoside and malvidin-3-
O-acetylglucoside. Coumaroyl-glucosides grouped petunidin-3-
O-coumaroylglucoside, peonidin-3-
O-coumaroylglucoside and malvidin-3-
O-coumaroylglucoside. At bottling time T1, the Lagrein red wine had the highest concentration of anthocyanins compared to Merlot and St. Magdalener. The total anthocyanins content of Lagrein was almost double that of Merlot, whereas non-acylated anthocyanins in Lagrein were about 265 mg L
−1 compared to about 160 mg L
−1 for Merlot and about 142 mg L
−1 in St. Magdalener. Acetylated and coumaroylated anthocyanins were also present in higher concentrations in Lagrein red (86 mg L
−1 and 27 mg L
−1, respectively), followed by Merlot (48 and 19 mg L
−1, respectively) and St. Magdalener (29 and 13 mg L
−1, respectively). Rosé wine produced from Lagrein had a predictably much lower concentration of anthocyanins (non-acylated, about 28 mg L
−1; acetylated, about 14 mg L
−1; coumaroylated about 7 mg L
−1) compared to the three red wines, due to the shorter maceration time. Nonetheless, in all four wines the sum of the three classes and thus the total amount of identified anthocyanins showed a clear decrease over the storage period. The reduction of the phenolic content during the bottle storage is generally ascribed to polymerization, oxidation and complexation reactions [
25,
26,
27]. In this case the dissolved oxygen present in the four wines (
Table 3) was low already just after bottling (0.15–1.7 mg L
−1 in red and 1.4–3.1 mg L
−1 in rosé wine) and varied slightly during 12 months of storage.
2.2. Effect of the Type of Stopper and Storage Time on the Phenolic Profile
Two-way ANOVA was performed to assess the influence of the type of stopper and the bottle storage time on the phenolic composition of the four wines. The results showed that the storage time (F storage time) significantly influenced the phenolic concentration (
Table 1 and
Table 2). Notably, GRP and anthocyanin glucosides, acetyl-glucosides and coumaroyl-glucosides in Lagrein rosé (F = 8570; 34,170; 20,423 and 10,927, respectively); caffeic acid, GRP and glucoside in Merlot (F = 3435; 3263 and 3435, respectively); GRP, glucosides and acetyl-glucosides in St. Magdalener (F = 2861; 4767 and 2971, respectively); and
p-coumaric acid in Lagrein (F = 4297) were strongly affected by the storage period. On the other hand, the influence of the type of stopper at each sampling time (F stopper) a significantly affected the phenolic compounds only in a few cases.
Compared to the bottles closed with conventional stoppers, Lagrein red wine closed with the ‘blend’ stopper showed significantly lower concentrations of anthocyanin glucosides at three months (T3), acetyl glucosides at six months (T4), (+)-catechin and caffeic acid at three and six months and p-coumaric acid at one month of storage (T2). Merlot wines reported significant differences for anthocyanin acetyl-glucosides at bottling (T1), at one month (T2) and at three months (T3), but in this case higher concentrations were detected in the bottles closed with the ‘blend’ stoppers compared to the conventional stoppers. All these differences can be attributable to an imperfect homogeneity of the Merlot wine in the bottling line (for each sampling point, the content of different bottles were analyzed).
The Lagrein rosé wine showed a higher abundance of p-coumaric acid at T2 in the wine sample closed with the ‘blend’ stopper. St. Magdalener wines differed for the lower concentrations of GRP, anthocyanin glucosides and acetyl-glucosides after six months storage (T4) in bottles closed with the ‘blend’ stoppers.
To summarize, a common temporal trend of one or more phenolic compounds related to the type of stopper could not be evidenced in the four wines. The statistical differences highlighted between two types of stoppers were quite unpredictable and seemed to be due to the natural variability of the products.
To better understand the influence of storage time and type of stopper on the phenolic profile in relation to dissolved oxygen, PCA was performed with all the phenolic compounds and the dissolved oxygen values obtained at each step of analysis. Rosé wines were excluded from the PCA because their content of phenolic compounds was lower than the red wines. The PCA plot showing the loading of variables and the scores of the samples in the bi-dimensional space is reported in
Figure 1.
The first two principal components made up 73% of the model (PC-1 42%, PC-2 31%) when taken together. Samples collected at bottling time (T1), after one month (T2) and after three months (T3) were grouped on the negative side of the PC-1 axis, whereas the wines stored for six months (T4) and 12 months (T5) were placed on the positive side of the PC-1 axis. Considering the loadings, the three classes of phenolic anthocyanins (glucosides, acetyl-glucosides and coumaroyl-glucosides) and non-anthocyanin phenolic compounds, such as syringic acid, showed a positive correlation with Lagrein red samples collected at bottling time (T1), three months (T2) and at six months (T3). On the other hand, catechin and epicatechin showed a good correlation with Santa Madgalener and Merlot samples analyzed at T1, T2 and T3. Gallic acid, protocatechuic acid, GRP, caffeic acid, p-coumaric and trans-resveratrol were positively correlated with the Merlot and Lagrein samples collected after six months (T4). Caftaric acid characterized all the wines stored for 12 months (T5), as well as the St. Magdalener wines sampled at six months storage (T4).
The dissolved oxygen was not strongly correlated with any specific sample.
This PCA model successfully scattered the samples depending on the storage period, whereas the bottles closed with similar stoppers were not clearly separated in any case. The multivariate approach confirmed the ANOVA results, remarking the dominant influence of the storage time over the type of stopper.
2.3. Evolution of Volatile Compounds during the Storage in Bottle
A total number of 26 volatile compounds were identified in all the wines (
Table 4).
The main goal of the SPME-GC/MS analysis was not to obtain the absolute concentrations of the volatile compounds present in the wine headspace, but to study the relationships between their relative concentrations (profiles) and the technological variables (wine type/stopper type/storage time) with a multivariate statistical approach.
The abundances of the volatile compounds (averages and standard deviations of two replicate bottles) were expressed as relative percentage (internal area) considering all the volatile compounds, as reported in
Supplementary Materials (tables of volatile compounds’ average peak areas in the studied wines).
The most represented compounds were short- and medium-chain esters, which included 17 compounds, followed by higher alcohols, short- and medium-chain organic acids and terpenes. During the bottle storage, more than 80% of the total abundance was represented by seven compounds: isopentyl acetate, 1-hexanol, ethyl hexanoate, 2-phenylethyl alcohol, diethyl succinate, ethyl octanoate and ethyl decanoate. Over 12 months, the modification of the volatile profile was primarily due to the evolution of these compounds. Ethyl octanoate and ethyl decanoate showed a decreasing trend in the early period of bottle storage (T2–T4), followed by an increase until the 12th month (T5) in red wines. Isopentyl acetate increased after one month (T2) of bottle storage in the three red wines and reached its maximum abundance at three months (T3), probably as a consequence of acid-catalyzed reactions involving fatty acid esters and resulting in the production of acetate esters [
38,
39].
The sum of higher alcohols decreased in all four of the wines. This reduction was detected in previous studies [
40,
41,
42] and can be explained as the result of the previously cited acid-catalyzed esterification reaction [
43]. Concerning the total amount of esters, Merlot is notable for having experienced a slight reduction during the 12 months in wine closed with a conventional stopper, whereas the Merlot, Lagrein rosé, Lagrein red and St. Magdalener wines closed with the ‘blend’ stopper showed a higher total ester content at the end of the storage compared to the start of the storage. This latter trend can occur as a consequence of the esterification of branched acids to form ethyl esters [
44,
45,
46].
In order to assess further relations between the type of stopper and the volatile compounds, PCA was applied to all of the red and rosé wine samples, the volatile compounds and the dissolved oxygen levels detected during the storage period. The loadings and variables of the PCA are displayed in
Figure 2.
The first two principal components explained 48% of the model (PC1, 29%, PC2, 19%). The PCA showed a separation of the rosé samples in the left part of the bi-plot with the T3 samples clustered away from all other samples along PC1. The red wines collected at bottling (T1) and after one month (T2) were grouped in the middle of the bi-plot, whereas samples stored for three months (T3) were segregated in the upper-right quadrant and samples collected after six months (T4) and 12 months (T5) were grouped together in the lower-right quadrant.
Some of the volatile compounds related to fresh fruity and floral notes, such as limonene (
12), ethyl octanoate (
21) and ethyl decanoate (
24) positively correlated with samples stored for short periods (T1 and T2). On the other hand, a volatile compound connected with oxidation reaction such as diethyl succinate (
18) distinguished samples stored for longer periods (T4 and T5). The rosé wine showed a positive correlation with 2-phenethyl acetate (
23) and hexyl acetate (
11); acetate esters are sensory descriptors of the fruity character of rosé wines [
47]. Also, this PCA scattered the samples depending on the storage period, whereas there was a lack of differentiation among the bottles closed with the two stoppers. This explains the similar influence of the different stoppers on the volatile profiles and on the dissolved oxygen level during 12 months of storage.
No consistent significant difference in terms of stopper or time-stopper interaction was evidenced by two-way ANOVA for any wine in terms of specific variables, although some compounds showed differences at specific times for stoppers. The main significance was shown in relation to time evolution, which involved most of the volatile compounds for all wines, as shown for the phenolic compounds. These will be discussed in more details in relation to the sensory test results.
2.5. Sensory Analysis
The results of the discriminant triangle test (
Table 5) showed significant differences for St. Magdalener and Merlot wines closed with the two different stoppers at the bottling time (T1), probably due to imperfect homogeneity of the wine in the bottling line. A significant number of assessors also recognized differences concerning Merlot wine in the tasting session performed after six months of storage (T4). Conversely, the required number of correct answers for differentiating the two samples closed with the different stoppers for any of the four wines was not obtained in the other tasting evaluations. An overall view of the tasting sessions highlights only few cases of significant differences. So, the differences were presumably due to the natural variability of the product.
Further investigations were performed to identify the compounds possibly responsible for the correct discrimination of St. Magdalener at T1 and Merlot wine at T1 and T4. A PCA model was built for St. Magdalener samples upon volatile compounds in order to identify the species most relevant for for the differentiation of the samples (
Figure 4).
The largest separation was identified between three groups of samples: T1-T2, T3 and T4, with T5 being more or less precisely at the center of the PCA model (considering PC1 and PC2). Two-way ANOVA was applied on all compounds to study the significance (p-value < 0.05) of the differences across the samples for specific variables, in particular with respect to the time evolution between stoppers and their interaction. Significant differences for the stopper were found for 2-ethylhexanol (13, p-value ~ 0.05), 1-octanol (15, p-value ~ 0.05). Significant differences for time evolution were found for acetic acid (1), 1-hexanol (5), isopentyl acetate (6), 2-butyl 4-ethylbenzoate (7), 1-heptanol (8), 1-octen-3-ol (9), hexyl acetate (11), limonene (12), 2-ethylhexanol (13), methyl benzaldehyde (14), 1-octanol (15), ethyl benzaldehyde (16), 2-phenylethanol (17), diethyl succinate (18), octanoic acid (19), methyl salicylate (20), ethyl octanoate (21), ethyl benzeneacetate (22), 2-phenyl acetate (23), ethyl decanoate (24), ethyl dodecanoate (25) and ethyl hexadecanoate (26). No significant interaction between the two factors was found.
Only borderline differences (at
p-value ~ 0.05) were observed for the type of stopper. ANOVA indicated that the only significant differences observed between the samples were occurring as a time evolution, and in agreement with the interpretation of PC1 and PC2 in
Figure 4. In fact, no variable was able to differentiate wines for the stoppers at a specific time.
As done with St. Magdalener, a PCA model was built on Merlot samples upon volatile compounds in order to identify the volatiles most differentiating the wines by the stopper (
Figure 5).
T1 and T2 samples were tightly clustered at negative values of PC1 and PC2. T3 samples were found at the highest positive values of PC1 and PC2, while at the same time completely separate from all other samples. T4 samples were found at the most negative values of PC2, whereas T5 samples were grouped in the same general direction, although much closer to the center of the PCA model with respect to both principal components. T4 samples showed again (as for St. Magdalener) to be between T3 and T5 with respect to PC1; similarly, the evolution from T3 to T5 appeared to opposed that from T1 to T3 along PC1. The separation of T4 from T3 and T5 occurred along PC2, but T3 samples were farther away than T5 in that direction, therefore indicating that these T4 samples could be regarded as a turning point in the wine evolution, with respect to the variables most relevant for PC2. Two-way ANOVA was applied to investigate the significance (p-value < 0.05) of differences for the time evolution, stoppers and possible factor interactions. As a result, significant differences (s.d.) were found with respect to the stopper for ethyl butanoate (2; p-value ~ 0.05), methyl benzaldehyde (14), 1-octanol (15), octanoic acid (19) and methyl salicylate (20).
Significant differences with respect to time were found for ethyl 2-methylbutanoate (
3), ethyl 3-methylbutanoate (
4), 1-hexanol (
5), isopentyl acetate (
6), 2-butyl 4-ethylbenzoate (
7), 1-heptanol (
8), ethyl hexanoate (
10), 2-ethylhexanol (
13), 4-methylbenzaldehyde (
14), 1-octanol (
15), 4-ethylbenzaldehyde (
16), diethyl succinate (
18), octanoic acid (
19), methyl salicylate (
20), ethyl octanoate (
21), ethyl benzeneacetate (
22), ethyl decanoate (
24), ethyl dodecanoate (
25) and ethyl tetradecanoate (
26). Significant interactions between time and stopper were found for ethyl butanoate (
2), 4-methylbenzaldehyde (
14), 1-octanol (
15), octanoic acid (
19) and methyl salicylate (
20). These were the same compounds differentiating the wines for stoppers. A Tukey’s HSD test highlighted how several differences were found for time between the two stoppers (hence the interaction), but also that all these compounds differentiated the wines at T3 between the two stoppers. The main differences were seen overtime. In particular, no specific compound could be indicated as responsible for the difference between the two stoppers at T4. Further, as seen in
Table 1 and
Table 2 and
Section 3, the phenolic compounds’ profiles (anthocyanins and non-anthocyanins) did not account for the differences observed in the triangle tests; instead, they mostly showed a dependence on storage time in the PCA (
Figure 1), with the largest differences at T3 for all wines, mostly due to storage time, with few exceptions. Therefore, further investigations (especially in terms of descriptive sensory analysis and quantitative sensory models) will be therefore required in order to be correlated to the parallelly studied chemical profiles.