1. Microorganisms

In this study, Hanseniaspora uvarum BHu9 and Saccharomyces cerevisiae BSc114, isolated from oenological environmental, were used. These yeasts have been molecularly identified in previous studies (Maturano et al., 2015a) and were selected according to their fermentative performance and respiratory characteristics, to be used in sequential inoculation, to obtain wines with reduced ethanol content (Mestre Furlani et al., 2017). The commercial yeasts EC1108 (Lallemand, Montreal, Canada) were employed as control. Three treatments were implemented and defined as fermentative strategies (FS): CI, co-inoculation of H. uvarum BHu9/S. cerevisiae BSc114 native yeasts; NS, S. cerevisiae BSc114 native yeast; and CS, S.&nbspcerevisiae EC 1118 commercial yeast. For CI treatment, BHu9 strain was inoculated at moment 0 in a concentration of 5*10⁶&nbspcells/mL, then 2*10⁶&nbspcells/mL of BSc114 was sequentially inoculated (after 48 h). This inoculation strategy was optimized and validated under laboratory conditions to reduce the ethanol content in wines (Maturano et al., 2019). For NS and CS treatments, the corresponding yeast strains were inoculated at time 0 with 2*10⁶&nbspcells/mL. Before the fermentative process, native yeasts inocula were prepared according to the procedure proposed in Maturano et al. (2019).

2. Winemaking procedure and experimental conditions

Grapes from two cultivars, Sauvignon blanc and Malbec (Vitis vinifera L.), were obtained from commercial vineyards located in Agrelo (68°51′W and 33°06′S), Luján de Cuyo, Mendoza, Argentina, during the 2017 season.

Sauvignon blanc grapes were manually collected at the beginning of ripening to obtain a juice with a very low sugar concentration (around 6.2&nbsp% v/v of potential alcohol degree) and a high acidity (>20&nbspg/L). One hundred kilograms of grapes were transported to the experimental winery of INTA (Mendoza, Argentina). Upon reception, were pressed in a manual screw press to obtain 60&nbspL of an unripe grape juice (115 g/L of sugar, 21.1&nbspg/L of titratable acidity, pH&nbsp2.53). The must was immediately sulphited (100 mg K₂S₂O₅/L), settled overnight at 4&nbsp°C, racked to a 50-L stainless steel tank, and inoculated with the commercial yeast EC1118 (Lallemand, Montreal, Canada). Alcoholic fermentation was carried out at 17 ± 2&nbsp°C. When finished, wine were sulphited (100 mg K₂S₂O₅/L) and kept at 2–4&nbsp°C until implementation in the blending assay with well-ripened Malbec grapes. This low-ethanol wine (LW) registered 6.5&nbsp% v/v of ethanol, titratable acidity of 19.5&nbspg tartaric acid/L, pH of 2.65, and the absence of colour and herbaceous aromas.

Subsequently, Malbec grapes were hand-picked at two different ripening stages. The first harvest (1H) was carried out when the potential degree of alcohol was approximately 13.0 % v/v (around 22&nbsp°Brix, 215&nbspg/L of sugar, 6.0&nbspg/L of titratable acidity, pH 3.40). The second harvest (2H) took place when the grapes reached optimum phenolic maturity (around 24&nbsp°Brix, 245&nbspg/L of sugar, 5.6&nbspg/L of titratable acidity, pH 3.82) with 14.5–15.0 % v/v of potential degree of alcohol. Grapes (1H, 225 kg; 2H, 450&nbspkg) were crushed, destemmed (Metal Liniers model MTL 12, Mendoza, Argentina), sulphited (100 mg K₂S₂O₅/kg), and the musts (skins, seeds, flesh and juice) placed into 25-L food-grade plastic tanks.

The experimental design consisted of three treatments with the selected yeasts, previously proposed, and applied by triplicate to the fruit of each harvest: 1H-CS, 1H-NS, 1H-CI, 2H-CS, 2H-NS and 2H-CI. Furthermore, from the second harvest, a part of the total volume of the grape juice was removed and replaced with the same volume of Sauvignon blanc low-ethanol wine (LW), as a strategy for alcohol reduction in wines (2RW). In this matrix, the different selected yeasts were also applied, obtaining the treatments 2RW-CS, 2RW-NS and 2RW-CI. This substitution volume was calculated, for each experimental unit, to reproduce the probable alcoholic degree of the corresponding wine from the first harvest (22&nbsp°Brix), following the equation proposed by Kontoudakis et al. (2011). This substitution represented around 20&nbsp% of the weight of the destemmed and crushed grapes.

Consequently, 27 vinifications [3 pre-fermentative treatments (PFS) × 3&nbspfermentative treatments (FS) × 3 replicates] were conducted at 25 ± 2&nbspºC, with a maceration length of 14&nbspdays. For cap management, two daily punch-downs (morning and afternoon, 1 min each) were applied. All tanks were daily controlled through the measurement of temperature and the weight loss of the fermenting systems. The monitoring of the growth of yeast populations was carried out by the periodically withdrawn samples from all fermentations. These samples were serially diluted, spread onto WLN-Agar medium, and incubated during 5–7&nbspdays at 28 ± 1&nbsp°C (Pallmann et al., 2001). Once fermentation-maceration was completed, free-run wines were collected into 10&nbspL glass carboys fitted with airlocks. Malolactic fermentation (MLF) was induced with a commercial Oenococcus oeni culture (VP-41, Lallemand, Montreal, Canada). After MLF finished, the wines were racked off the lees, adjusted to 35&nbspmg/L of free SO₂, and stored at 1–3&nbsp°C for 30&nbspdays to allow tartaric stabilization. Finally, wines were bottled and stored in a dark cellar at 12–15&nbsp°C until analysis. In all cases, the analyses were completed in about two months, starting from the second month after bottling.

3. Grape and wine general analytical parameters

For grape analysis, one hundred berries were randomly selected, from each cultivar and corresponding harvest time, and used to measure the sugar concentration, pH, and titratable acidity (OIV, 2012). For wine analysis, standard parameters including titratable acidity (tartaric acid, g/L), volatile acidity (acetic acid, g/L), pH, residual sugar (g/L), and alcohol content (% v/v) were determined as described by International Organization of Vine and Wine (2012), using an ALPHA FT-IR Wine Analyzer (Bruker Optics, Ettlingen, Germany), and alcohol tester (Alcolyzer Wine®; Anton-Paar GmbH, Austria). The ethanol yield metabolic parameter was calculated from the following analytical data: the ratio between sugar required (initial sugars - final sugars) in g/L and the percentage of ethanol produced (%&nbspv/v).

4. Phenolic composition and colour parameters

Wine samples were centrifuged (11,000&nbspg × 5&nbspmin) and filtered through with 0.22-mm membranes (Microclar, Buenos Aires, Argentina) before analysis. Absorbance measurements were made with a Perkin-Elmer UV–visible Spectrophotometer Model Lambda 25 (PerkinElmer, Hartford, CT).

Tannins were analysed by protein precipitation (Harbertson et al., 2003). Anthocyanins, small polymeric pigments (SPP), large polymeric pigments (LPP), and total polymeric pigments (SPP + LPP) were measured as previously described (Harbertson et al., 2003). Iron reactive phenolics (total phenols) were analysed following the method described by Heredia et al. (2006).

CIELAB parameters [L*(lightness, 0&nbspblack and 100&nbspwhite), C*_ab (chroma, saturation), h_ab (tone; red, green, yellow) and the a*b* (red/green; yellow/blue) coordinates], were calculated from the absorption spectra by using the CromaLab® software (Heredia et al., 2004), following the recommendations of the Commission Internationale de L'Eclairage (CIE, 2004). Colour difference (ΔE*_ab) was calculated as the Euclidean distance between two points (1 and 2) in three-dimensional (L*a*b*) space. ΔE*_ab (L*₁, a*₁, b*₁; L*₂, a*₂, b*₂) = [(ΔL*)² + (Δa*)² + (Δb*)²]^1/2, where ΔL*=&nbspL*₁-L*₂; Δa*= a*₁&nbsp-&nbspa*₂ and Δb*=&nbspb*₁-b*₂.

5. Anthocyanins and derived pigments profile

The chromatographic system employed for wine anthocyanin identification and quantification was a Perkin-Elmer Series 200 high-performance liquid chromatograph equipped with a diode array detector, a quaternary pump, and an autosampler (HPLC-DAD; PerkinElmer, Shelton, CT). Separation was performed on a reversed-phase Chromolith Performance C18&nbspcolumn (100&nbspmm × 4.6&nbspmm&nbspI.D., 2&nbspμm; Merck, Darmstadt, Germany) with a Chromolith guard cartridge (10&nbspmm × 4.6&nbspmm) at 25&nbsp°C. A gradient consisting of solvent A (water/formic acid, 90:10,&nbspv/v) and solvent&nbspB (acetonitrile) was applied at a flow rate of 1.1 mL/min from 0 to 22&nbspmin and 1.5 mL/min from 22 to 35&nbspmin as follows: 96–85&nbsp% A and 4–15&nbsp%&nbspB from 0 to 12&nbspmin, 85–85&nbsp%&nbspA and 15–15&nbsp%&nbspB from 12 to 22&nbspmin, 85–70&nbsp%&nbspA and 15–30&nbsp%&nbspB from 22 to 35&nbspmin; followed by a final wash with 100&nbsp% methanol and re-equilibration of the column. Two millilitres of wine samples were filtered (0.45-μm pore size nylon membrane; Microclar, Buenos Aires, Argentina), and then 100&nbspμL-aliquot was injected onto the column. Diode array detection was performed from 210 to 600&nbspnm, and the quantification was carried out by peak area measurements at 520&nbspnm. Anthocyanin amount was expressed by using malvidin-3-glucoside chloride as standard for a calibration curve (R²&nbsp=&nbsp0.99). Identification and confirmation of anthocyanin pigments were performed by HPLC-DAD/ESI-MS as described by Blanco-Vega et al. (2011).

6. Sensory analysis

All the wines obtained were evaluated three months after bottling by a trained panel of 12&nbspvolunteers (8&nbspmales and 4&nbspfemales) aged 26–48&nbspyears, including researchers and technicians (EEA Mendoza INTA) with the previous wine tasting experience. To determine if wine replicates from each treatment were perceptibly different, we performed first discrimination tests (Triangle tests). Because no significant differences were found between the replicates, we performed a descriptive sensory analysis (DSA) with nine wines, one of each treatment (Heymann and Lawless, 2010). Consensus terminology, reference standards (Table 1), and scale practice were developed over four training sessions (of 1-hour each) held over two weeks. Panellists selected two colour attributes (colour saturation and violet hue), four taste and mouthfeel descriptors (acidity, bitterness, astringency, and fullness) and five aromas (fruity, floral, herbaceous, spicy, and balsamic). Following training, panellists were required to evaluate the nine wines by triplicate during three tasting sessions. Each session began with the assessment of the aroma standards. The William Latin Square design was used to control for carryover effects. Approximately 30–40&nbspmL of wine was served, at 16–18&nbsp°C, in clear wine tasting glasses (ISO 3591, 1977) labelled with three-digit code. For each descriptor, panellists had to rate the intensity of each wine on a scale from 1 to 10. To collect data from the panellists we used Smartphones with SOLDESA software (ISETA, Buenos Aires, Argentina). Panel performance was monitored by assessing the correlation of panellists with the panel mean and by their contribution to the panellist × wine interaction for each attribute. No information about the nature of the study was provided to reduce bias.

6. Data analysis

All chemical analyses were carried out in duplicate. Statistical analysis was assessed with Statgraphics Centurion XVI software (Statistical Graphics Corp., Warrenton, VA, 2009) and R (R&nbspCore, 2018). All results (chemical and sensory) were tested for homogeneity of variance using Cochran's test, and analysed by two-way analysis of variance (ANOVA) and Tukey's honestly significant difference (HSD) test (α&nbsp=&nbsp0.05). A&nbspp&nbsp<&nbsp0.05 was considered to be statistically significant. Principal component analysis (PCA) was performed on all the sensory attributes using the correlation matrix. Confidence ellipses indicating 95&nbsp% confidence intervals were based on the multivariate distribution of the Hotelling's test for p&nbsp<&nbsp0.05 and were constructed using SensoMineR panellipse function on R (Husson et al., 2005).

1. Yeast growth and fermentative kinetic

As to be shown in previous work (Mestre Furlani et al., 2017), both native yeasts, S.&nbspcerevisiae BSc114, and H.&nbspuvarum BHu9 tolerate high concentrations of SO₂. These yeasts started without difficulty the fermentative process at SO₂ levels applied at grape must, as can be observed in the short phase lag (Figure&nbsp1). Wild yeasts were detected the first stage of all trials, but at significantly lower proportions than inoculated starter yeasts. Resistance degree to SO₂ is variable at strain level and several non-Saccharomyces yeast species can be tolerant at commonly employed concentrations in winemaking (Mendoza et al., 2009; Padilla et al., 2016).

As expected, the fermentation times were closely related with the sugar concentration of the grape musts, that is the treatments that started the process with around 22&nbsp°Brix (1H-CS, 1H-NS, 1H-CI, 2RW-CS, 2RW-NS and 2RW-CI) finished the fermentative process in a range of 6–9&nbspdays. While 2H-CS, 2H-NS and 2H-CI fermentations were completed in 8–11&nbspdays (24&nbsp°Brix, initial grape must). In all cases, the treatments that included sequential co-inoculation, 1H-CI, 2H-CI and 2RW-CI, lasted more time demonstrating the lower fermentative ability of non-Saccharomyces yeast (Figure&nbsp1). Others authors employed co-cultures and obtaining similar results concluding that longer fermentations reduce the flavour lost or high energetic demand for refrigeration (Lleixà et al., 2016).

During the fermentation process, it was evident that all experimental units inoculated with the BSc114 yeast at 48&nbsph (CI&nbsptreatments) increased significantly your population size and CO₂ production. Furthermore, these exhibited the highest CO₂ production rates and population sizes during the first 4–5&nbspdays reaching maximum populations above 8&nbspLog CFU/mL and then slowly declined as fermentation progressed to completion (Figure&nbsp1c,f,i). The results of population dynamics indicated that S. cerevisiae BSc114 was able to overcome the autochthonous microbiota of the grape musts from the beginning. All these facts denote that this native yeast achieved rapid implantation.

Hanseniaspora uvarum is one of the predominant species at the beginning of the process and has been frequently reported by others authors (Lleixà et al., 2016; Tristezza et al., 2016), therefore sequential co-inoculation with S. cerevisiae is one of the most feasible strategies to mimic what happens during spontaneous fermentation and to improve the quality of the wine. In this study, all co-cultures registered very low fermentative activities during the period before the BSc114 yeasts inoculation compared with the pure cultures of S.&nbspcerevisiae (Figure&nbsp1). Other authors confirmed an initial reduced fermentative rate in&nbspH.&nbspuvarum/S.&nbspcerevisiae co-cultures (Ciani et al., 2006; Du Plessis et al., 2019; Tristezza et al., 2016). Despite the low fermentation activity, consumption of reducing sugars was 23.5&nbsp% (1H-CI), 20.4&nbsp% (2H-CI) and 34.2&nbsp% (2RW-CI) during this period. The highest consumption of sugars was detected in 2RW-CI, that can be attributed to the acidity affinity of the grape juice (pH&nbsp3.6, 6.5&nbspg/L titratable acidity) by this yeast, as reported by other authors (Çelik et al., 2017; Ciani et al., 2006). This sugar consumption was reflected in the increase in biomass favoured by the oxygen levels still present, as can be seen in Figure 1. From the beginning, H. uvarum BHu9 was present at populations higher than 6.0&nbspLog UFC/mL; it reached a maximum (between 6.86 and 7.73 Log&nbspCFU/mL) in three days and then decreased. This apiculate yeast´s behaviour was previously reported, which indicates that non-Saccharomyces yeasts dominate during the first days of fermentation up to an ethanol concentration of about 4–7&nbsp% v/v, and then they decline your growth (Fleet, 2003). BHu9 yeast population remained until Day 4 (1H-CI) and Day 6 (2H-CI and 2RW-CI). On the other hand, it is a relevant highlight that this early consumption diminishes the fermentable sugars for the S.&nbspcerevisiae starter. Therefore, it could be inferred that it reduced the impact of osmotic stress on this starter yeast. Furthermore, it no presented lag phase after its inoculation (at 48 h) and rapidly started to increase your population and CO₂ production. It is important to emphasize that the total populations reached by the co-cultures did not significantly exceed those found in vinifications conducted by BSc114 under pure conditions (data not shown).

Fermentation assays of the commercial yeast EC1118 (1H-CS, 2H-CS and 2RW-CS) always reached population sizes lower than fermentation with native yeasts. Furthermore, fermentations conducted by EC1118 registered CO₂ production higher than native yeasts (Figure&nbsp1). These results could indicate that the strain EC1118 presented a high sugar/ethanol conversion, that is, it requires less sugar to produce more ethanol.

Figure 1. Evolution of the yeast populations (Log CFU/mL) and CO2 production (g/L) during fermentation of Malbec wines obtained applying pre-fermentative and fermentative strategies of winemaking: (a) 1H-CS, (b) 1H-NS, (c) 1H-CI, (d) 2H-CS, (e) 2H-NS, (f) 2H-CI, (g) 2RW-CS, (h) 2RW-NS and (i) 2RW-CI.

Save View full size Expand inline Collapse inline

2. General analytical parameters of Malbec wines

The sequential inoculation of H. uvarum/S. cerevisiae yeasts was successfully implemented by researchers in others oenological regions, achieving good quality wines (Canonico et al., 2016; Ciani et al., 2006; Mendoza et al., 2009; Tristeza et al., 2016). The co-inoculation strategy applied in the present work was previously optimized and it achieved promising results on a laboratory scale (Maturano et al., 2019; Mestre et al., 2019). It is important to note that all the wines from the different treatments fermented to dryness (<&nbsp3&nbspg/L of total residual sugars, Table 2).

The final amounts of ethanol, pH, titratable and volatile acidity of the wines obtained are summarized in Table&nbsp2. As expected, the ethanol content of the wines elaborated from grapes harvested at technological maturity (2H) was significantly higher than wines from the first harvest (1H) and those made blending ripe must with a proportion of low-ethanol wine (2RW). Overall, regardless of the yeast strain used (FS), 2RW wines had 1.4&nbsp% v/v less alcohol than 2H&nbspwines, and levels close to 1H&nbspwines. Ethanol yield was also considered an important factor. As can be observed in Table&nbsp2, the fermentative strategy 2RW–CI registered the highest values, demonstrating its low efficiency in the conversion of sugars into ethanol. Likewise, the pH values of the wines showed similar behaviour, achieving a reduction of 0.3&nbspunits by applying the 2RW treatments; while the titratable acidity only increased by 4&nbsp%. Conversely, comparable pH and acidity levels were obtained between H1 and H2 wines; probably due to high potassium levels observed in grapes since the beginning of ripening (H1, 1500&nbspmg/L; H2, 1700&nbspmg/L).

These results are in agreement with previous experiments (Kontoudakis et al., 2011; Piccardo et al., 2019; Schelezki et al., 2018), in which the substitution of the grape juice with unripe must/wine determined a decrease in sugar content and an increase in acidity. Specifically, a huge effect of the pre-fermentative factor (60.2&nbsp% and 74.3&nbsp%) and a slight PFS x FS interaction effect (19.0&nbsp% and 9.0&nbsp%) on alcohol and pH, respectively, was observed. While, in the case of titratable acidity, the impact of the fermentative factor (PF) was greater than the rest (51.5&nbsp%). Therefore, the 2RW-CI and 2RW-NS combinations were the most effective treatments in reducing alcohol and pH of wines (Table&nbsp2).

As mentioned previously, the active participation of the non-Saccharomyces native yeast as initial inoculum influences the fermentation course and the quality of the final product. Some species of non-Saccharomyces yeasts are recognized as potential acidifying, such as Lanchacea thermotolerants, Schizosacchamyces pombe, Starmerella Bacillaris and Candida zemplinina (Benito et al., 2015; Gobbi et al., 2013). These yeasts are capable of inducing biological acidification due to their physiological features and genetic determinants associated with the production of organic acids (Berbegal et al., 2019). Hong and Park (2013) reported to H.&nbspuvarum strains as a good producer of organics acids in Korean wines.

Finally, while the volatile acidity increased with the fruit ripeness levels, all treatments presented acetic acid values lower to 0.6&nbspg/L (Table&nbsp2), considered acceptable according to the regulations in force (Instituto Nacional de Vitivinicultura, Argentina).

3. Global phenolic parameters of wines

Figure 2 shows the global phenolic parameters of Malbec wines obtained applying pre-fermentative and fermentative strategies of winemaking to reduce alcohol and pH. First, it is well known that grape maturity significantly influences on colour and phenolic composition of red wines (Bindon et al., 2013). Wines from the riper grapes (2H) presented higher total concentrations of phenols (~47&nbsp%), anthocyanins (~26&nbsp%), and to a lesser extent of tannins (~19&nbsp%) than their corresponding first harvest wines (1H), confirming trends previously observed for this variety (Fanzone et al., 2011) and other red cultivars (Fournand et al., 2006). The main differences were observed in total phenols, measured in reaction with iron chloride, counting all phenolics containing vicinal dihydroxyls (Harbertson and Spayd, 2006). Therefore, these measurements include not only tannins but also flavan-3-ols and flavonols, compounds that increase their concentration and possibly extractability during grape ripening (Fanzone et al., 2011; Liang et al., 2012).

However, when each reduced alcohol wine (2RW) were compared with controls (2H), no significant differences in these parameters were found, considered as satisfactory results concerning wine quality (Figure&nbsp2). These findings agree with those obtained in similar experiments carried out by other authors (Kontoudakis et al., 2011; Piccardo et al., 2019; Rolle et al., 2018; Schelezki et al., 2018; Schelezki et al., 2020), and confirmed that changes (reduction) in alcohol levels (around 1.5–2&nbsp% v/v) in the wine matrix did not affect the extractability of anthocyanins and tannins from the skins. Furthermore, the replacement of some of the juice, before maceration, by S. blanc low-ethanol wine did not involve losses of anthocyanins extracted during the crushing process, but even a possible supply of copigments, which protects anthocyanins against oxidation (Boulton, 2001). These factors must be added to low pH values, which slow down this reaction and favour the extraction of these compounds during winemaking (Canals et al., 2005; Kontoudakis et al., 2011).

Table&nbsp3 shows the probability values of each factor, when the two-way ANOVA was applied using "pre-fermentative (PFS)" and "fermentative (FS)" strategies as factors. From this analysis, a significant effect of the FS factor (yeast strain used) was also observed, indicating in all cases higher levels of tannins (~34&nbsp%) and anthocyanins (~12&nbsp%) in wines fermented with native yeasts (S.&nbspcerevisiae BSc114, NS) than those fermented with the commercial yeast (CS) (Figure&nbsp2). The higher extraction of these compounds could be explained by a positive effect of the lower pH levels obtained with the native strain (Table&nbsp2). It is well known that the yeast´s polysaccharides retain anthocyanins and tannins, thereby preventing their precipitation. Therefore, colour stability is improved and astringency diminished (Escot et al., 2001). These molecules are released from the yeast cell wall during alcoholic fermentation and wine ageing processes. The amount released is intraspecific and depends on the number of cells formed and their physiological conditions (Domizio et al., 2014). In this study, as can be seen in Figure&nbsp1, the native S. cerevisiae yeast reached high populations throughout the fermentation process, therefore a higher concentration of polysaccharides could be expected, although this was not determined in this work.

Polymeric pigments formed during winemaking provide stable colour and positive mouthfeel properties. Their production is modulated by the relative molar amount of anthocyanins and tannins (Kilmister et al., 2014). Total polymeric pigments result from the summation of small polymeric pigments (SPP; including pyranoanthocyanins and flavanol-anthocyanin ethyl-bridged adducts, that cannot precipitate proteins and are resistant to SO₂ bleaching) and large polymeric pigments (LPP; covalent adducts between tannins and anthocyanins, of relatively high molecular weight, that can precipitate proteins and are resistant to SO₂ bleaching) (Adams et al., 2004). The total polymeric content (TPP) of the wines made from riper grapes (2H and 2RW) was higher than wines from less mature grapes (1H). LPP formation followed the trend observed regarding the tannin content in wines, while the rate of SPP showed a direct correlation with the level of anthocyanins (Figure 2). Similar results were reported by other authors in Syrah (Garrido-Bañuelos et al., 2019), Cabernet-Sauvignon (Bindon et al., 2013) and Merlot wines (Casassa et al., 2019) obtained from full-ripe fruit harvested between 23 and 25&nbsp°Brix.

Two-factor ANOVA showed a significant pre-fermentative × fermentative treatments interaction for LPP and SPP (Table&nbsp3). This suggests that the effect of yeast strains employed (FS) did depend upon pre-fermentative strategy. In other words, the native strains (NS and CI treatments) applied to control wines from ripe grapes (2H) and reduced alcohol wines (2RW) produced contrasting effects on polymeric pigments of different molecular weight, without affecting the summation (TPP).

Figure 2. Global phenolic parameters of Malbec wines obtained applying pre-fermentative and fermentative strategies of winemaking.

Save View full size Expand inline Collapse inline

Different letters for each group of bars indicate significant differences (Tukey HSD test, p&nbsp<&nbsp0.05).

^a Considered to be significant when p&nbsp<&nbsp0.05.
^b Interaction effect between pre-fermentative (PFS) and fermentative (FS) strategies.

4. Wine colour

For further understanding of the colour composition of wines, we analysed them by the CIELAB colour space. Figure 3 shows the location of wines from the different treatments in the colour plane (a*b*) of CIELAB space and L* (lightness) values. Data obtained from the two-way ANOVA (Table&nbsp3), show a significant effect (p&nbsp<&nbsp0.05) of pre-fermentative and fermentative strategies in the final colour of Malbec wines, in both quantitative (C*_ab and&nbspL*) and qualitative (h_ab) terms; and significant interaction PFS × FS in chroma (C*_ab) and hue (h_ab).

Regardless of the yeast strain used, it was found that the colour of first harvest wines (1H) was less vivid and lighter (lowest C*_ab), and showed a slight decrease of violet hues, given a markedly decrease of a* and b* values. Conversely, the colour of the wines from riper grapes displayed high intensity due to an increase of C*_ab (2H, 47 %; 2RW, 78 %), and a shift towards a violet hue (decrease in h_ab of 5° and 8° for 2H and 2RW, respectively) (Figure 3A). Furthermore, C*_ab presents a negative correlation with L* and, as expected, when the former increased, the latter decreased (Figure&nbsp3B). The L* value is the vertical axis and defines the lightness, the property according to which each colour can be considered as equivalent to a member of the greyscale, between black and white, taking values within the range of 0−100, respectively (Gordillo et al., 2014). This parameter showed significant differences between wines, presenting a relative percentage decrease (on average) of 16&nbsp% in 2H and 19&nbsp% in 2RW, in relation to 1H wines. These results can be explained by the higher levels of phenolic compounds, especially anthocyanins and polymeric pigments, in 2H and 2RW wines (Figure&nbsp2).

Additionally, when each reduced alcohol wine (2RW) were compared with controls (2H), significant differences were also found in these parameters, 2RW wines had a deeper colour (higher&nbspC*_ab, ~20&nbsp%; lower&nbspL*, ~4&nbsp%). The determinant factor possibly is the lower pH, which displaces the balance between the different forms of the anthocyanins towards the flavylium cation. Once again, these findings coincide with those obtained by other authors (Kontoudakis et al., 2011; Piccardo et al., 2019) in similar trials.

As mentioned above, it was observed a significant influence of the yeast strains and the PFS&nbsp×&nbspFS interaction on the colour parameters of wines (Table&nbsp3). The inoculation of S. cerevisiae native strain (BSc114), especially as a single inoculum (NS), for each pre-fermentative treatment, generated greater colour intensity in wines (higher&nbspC*_ab), with a tendency towards violet-blue hues (lower&nbsph_ab) (Figure&nbsp3A). Likewise, the L*&nbspvalues in NS wines were lower than those of CI and CS treatments, for each pre-fermentative strategy (Figure&nbsp3B). These results are consistent with the higher levels of anthocyanins and tannins in the wines obtained with this native yeast (Figure&nbsp2).

Finally, we also evaluated the total colour difference (ΔE*_ab) between Malbec wines obtained by applying different pre-fermentative treatments and fermented with different yeast strains. It is not possible to establish the concrete value for the colour discrimination because many factors are conditioning this limit, such as viewing geometry, surroundings, even the colour region and the ability, visual capacity, and training of observers. Some studies performed under specific conditions have established the visual discrimination threshold in wines between 3 and 5 CIELAB units (Martínez et al.,&nbsp2001, Pérez-Magariño and González-SanJosé, 2003), although these values must be considered in relative character. For all yeast studied, the highest colour difference values (ΔE*_ab) were found between first harvest wines (1H) and reduced alcohol wines (2RW) (mean 23.6 CIELAB units), and the least ones between second harvest control wines (2H) and 2RW wines (mean 7.6&nbspu), although in all cases they were visually distinguished (Figure&nbsp4). Taking into account that both 2H and 2RW wines come from grapes with the same degree of ripeness, the implementation of the alcohol reduction strategy significantly changed the colour, mainly in quantitative terms (%Δ²C*_ab&nbsp>&nbsp82).

Analysing the influence of yeast strains, the native inoculum applied (S. cerevisiae BSc114,&nbspNS) generated the highest colour difference (mean 18.0&nbspu) in the wines obtained from all pre-fermentative treatments (PFS). While the commercial strain (CS) showed the least colour differences (mean 13.8&nbspu) in all the wines. Following the same trend indicated above, the microbiological strategy of alcohol reduction applied also modified the colour of the wines in mainly quantitative terms (%Δ²C*_ab&nbsp>&nbsp60) (Figure&nbsp4).

Figure 3. CIELAB parameters of Malbec wines. (A) (a*b*) Diagram; chroma (C*ab) is a vector that links the dot (wine location) with the origin of coordinates; hue (hab) is the angle of this vector. (B) Lightness (L*).

Save View full size Expand inline Collapse inline

Figure 4. Colour differences (ΔE*ab), with the relative contribution of lightness, chroma and hue (%Δ2L, %Δ2C, %Δ2H), between Malbec wines obtained applying different pre-fermentative treatments and fermented with different yeast strains.

Save View full size Expand inline Collapse inline

5. Anthocyanin profile of Malbec wines

The identified and quantified compounds in the wine samples are summarizes in Tables S1-S5 (Supplementary data). They were grouped in non-acylated glucosides (5), acetyl-glucosides (5), cinnamoyl-glucosides (6), and low molecular anthocyanin-derived pigments (8 pyranoanthocyanins, and 2 flavanol-anthocyanin adducts). All identified compounds were detected in all of the wines studied. Figure&nbsp5 shows the results obtained in wines grouped by families. The simple glucosides represented the highest proportion of all anthocyanins in the samples (68.8&nbsp%), followed by acetylglucosides (15.8&nbsp%), cinnamoylglucosides (7.4&nbsp%), and pyranoanthocyanins (6.2&nbsp%), in accordance with the results previously published by our group for the same cultivar (Fanzone et al., 2012).

The overall amount of anthocyanins were similar to those obtained by spectrophotometry (Figure&nbsp2), although the values were significantly lower. However, the increasing trend detected between the first (1H) and second harvest (2H), by the mentioned technique, was not observed in the individual quantification by HPLC-DAD, presenting similar levels, mainly in the malvidin derivatives. The only compounds that increased at the second harvest (2H) were petunidin-3-glucoside, peonidin-3-glucoside, delphinidin-3-(6''-acetyl)-glucoside, cyanidin-3-(6''-acetyl)-glucoside, delphinidin-3-(6''-p-coumaroyl)-glucoside, and malvidin-3-(6''-caffeoyl)-glucoside (Tables S1-S3, supplementary data). This behaviour could be explained by the fact that HPLC-DAD analysis only detects free anthocyanins, whereas spectrophotometric analysis overestimates their total amount including other red pigments (Canals et al., 2008). Furthermore, the substitution of part of the juice of ripe grapes by low-alcohol wine (2RW) did not imply losses of anthocyanins, coinciding with the results presented above (Figure&nbsp2).

It should be noted that the concentration of pyranoanthocyanins and flavanol-anthocyanin adducts in the wines was significantly affected by the prefermentative strategy applied, and in the case of vitisins also by the yeasts used (Figure&nbsp5, Table&nbsp4). In addition, data obtained from the two-way ANOVA, show a significant interaction PFS × FS (p <&nbsp0.05) only for hydroxyphenyl derivatives. Pyranoanthocyanins are anthocyanin-derived pigments formed by the reaction of monomeric anthocyanins with acetaldehyde, acetoacetic acid, pyruvic acid and other carbonyl compounds (Blanco-Vega et al., 2011; Rentzsch et al., 2007). These compounds are of interest for winemakers because they have high stability during the ageing of red wines, are more resistant to elevated pH values and bisulphite bleaching than anthocyanins, and express more colour than other pigments at the typical pH of wine (Rentzsch et al., 2007). Meanwhile, ethylene-bridged flavanol-anthocyanin adducts often disappear during wine ageing, maybe due to the reported instability of the ethylidene linkage (Escribano-Bailón et al., 2001).

In our study, the Malbec wines from ripe grapes (2H and 2RW) fermented with native or commercial yeasts (NS and CS) contained the highest levels of all derived pigments, which support the greater proportion of small polymeric pigments obtained by spectrophotometry (Figure&nbsp2).

Figure 5. Anthocyanins and derived pigments profile of Malbec wines obtained applying pre-fermentative and fermentative strategies of winemaking.

Save View full size Expand inline Collapse inline

Different letters for each group of bars indicate significant differences by two-way ANOVA (Tukey HSD test, p&nbsp<&nbsp0.05).

^aConsidered to be significant when p < 0.05.
^bInteraction effect between pre-fermentative (PFS) and fermentative (FS) strategies.

6. Descriptive sensory analysis

Descriptive analysis with a trained panel was conducted to determine the specific sensory effects of both pre-fermentative and fermentative strategies. A fixed-effect two-way ANOVA with interaction was performed on the sensory data (Table&nbspS6, supplementary data). There was no interaction between pre-fermentative and fermentative treatments in any of the wine’s attributes, showing there were not any synergistic or antagonistic sensory effects of both strategies. The effect of the pre-fermentative strategy prevailed in all the sensory attributes. In brief, the descriptive analysis showed significant pre-fermentative effects for colour saturation, violet hue, fruity, spicy, balsamic, bitterness, and fullness. Although there were chemical differences in total acidity, these differences were not perceived by the sensory panel. 2H and 2RW wines showed higher colour saturation and violet hue than 1H wines. Contrary to the results obtained in CIELAB parameters (Figure&nbsp3), there were no significant differences in visual appreciation of colour saturation when reduced alcohol wines (2RW) were compared with controls (2H). This result provided evidence that the use of low-alcohol wine in the Malbec wine matrix did not negatively affect the perception of colour.

The sensory aromatic profile of the wines revealed that the main differences were found between 1H wines and both treatments coming from the second harvest. According to the DSA, producing Malbec wines that were up to 1.4&nbsp% v/v less alcohol by employing reduced alcohol wine (2RW) did not cause significant changes in the perceptions of any aroma when comparing with wines from the same harvest time (2H). Schelezki et al. (2020) reported similar results in Cabernet Sauvignon and Shiraz wines from Australia, and Piccardo et al. (2019) in Pinot noir and Tannat wines from Uruguay.

The wines behaved in a similar way for fullness and astringency attributes, being both wines from second harvest (2H and 2RW) the ones with a higher intensity. In the case of astringency, there was also a significant fermentative effect, CS wines showed fewer levels of astringency than NS and CI. In both cases, the perception of this mouthfeel attribute was in agreement with the measurement of tannin content by protein precipitation showed in Figure 2. There is considerable evidence (Kutyna et al., 2010) to support that wines with high alcohol content can be more bitter than control ones. This situation is evident in our experiment in which 2H wine is bitterer than the others are.

To further explore the comparative influence of both pre-fermentative and fermentative strategies on the sensory profile of the wines, the full dataset was submitted to principal component analysis (PCA). The PCA biplot and confidence ellipses were constructed with 95&nbsp% certainty according to the Hotelling's test (Husson et al., 2005), which provides significance testing. The sizes of the confidence ellipses are related to the variability of each wine (Figure&nbsp6A), while the colour dots in the loading plot showed the variability around the sensory attributes (Figure&nbsp6B). Dimensions 1 showed the wines were separated mainly by pre-fermentative strategies, with the first principal components explaining ~63&nbsp% of the total variance (Figure&nbsp6). The 1H wines clustered in the negative region of dimension 1, whereas 2H and 2RW wines grouped on the positive region of dimension 1 (Figure&nbsp6A). The factor map showed significant overlap between wines from the first harvest time (1H) indicating no differences between fermentative strategies applied to that matrix. A similar situation occurred with the ellipses of wines 2H and 2RW wines except for wines 2H-NS and 2RW-CI that appear to be significantly different wines. The corresponding loading plot (Figure&nbsp6B) showed the separation of the wines in the first principal components could be attributed to the herbaceous descriptor negatively loaded in the first dimension with almost the rest of the attributes positively loaded. The exception was acidity and bitterness characters that were positively and negatively loaded in the second dimension.

Figure 6. Principal component analysis (PCA) of descriptive sensory data of Malbec wines evaluated by a trained panel (n = 12). (A) wine factor map. (B) sensory attributes loadings with 95 % confidence ellipses based on the multivariate distribution of Hotelling’s test (p < 0.05).

Save View full size Expand inline Collapse inline