Malvasia aromatica white grapes cultivated in Madrid winegrowing region were hand-harvested and elaborated in the Experimental Winery from IMIDRA Institute at the “Finca El Encín”, in Alcalá de Henares, Spain.

The S. cerevisiae yeast strains (CLI 271 and CLI 889) were isolated from wineries of the PDO “Vinos de Madrid” and selected based on its good oenological properties by the Oenological Microbiology Laboratory of the Department of Agrifood Research at Madrid Institute for Rural, Agriculture and Food Research and Development (IMIDRA) Institute. These two native strains are well-known and widely tested at IMIDRA laboratories. Specifically, S. cerevisiae CLI 889 presented a high fermentative capacity and optimal implantation rate, resistance to ethanol, fruity and fresh character and it is low-producer of acetic acid, hydrogen sulfide and sulfur dioxide [19, 20]. Likewise, the use of CLI 271 strain was interesting for its capacity of consumption of amino acids precursor of oxidation notes, reducing oxidation notes after sensorial analysis [21]. These strains were tested versus a S. cerevisiae active dry commercial yeast (LSA Vario from Agrovin), considered as control. The fermentations were performed on must previously macerated at 10 °C for 18 h. After cold maceration, Malvasia aromatica must showed 21.4 °Brix and 164 mg/L of yeast assimilable nitrogen (YAN), the pH value was 3.32 and the titratable acidity was 5.4 (expressed as g/L of tartaric acid). The must was divided into seven tanks. Moreover, the addition of nutrients (Actimax plus 30 g/hL) was done at the beginning and half of fermentation to encourage the growth of yeasts.

Three were inoculated with the CLI 271 yeast, three others with the CLI 889 strain and the last one with the commercial yeast (LSA, was used as a control). Each S. cerevisiae strain was inoculated in grape must at a concentration of 10${}^6$ cells/mL, from a pre-culture grown for 48 h in Yeast extract Peptone Dextrose (YPD) liquid medium at 28 °C. Fermentation took place under controlled temperature of 16 °C and was followed daily by measuring density (Proton 20969 and 28271 densimeters, Spain).

Pre- and post-fermentation parameters were analyzed from each treatment replication. International Organisation of Vine and Wine (OIV) official methods [22] were used for the analysis of °Brix, free and total sulphur dioxide, pH, titratable acidity, residual sugars and volatile acidity. Fermentation kinetic was controlled by daily monitoring of the density. The fermentative capacity was valuated as the difference between the initial and the final sugar content.

Twenty-six major volatiles were determined by gas chromatography coupled to flame ionization detector (GC-FID; Agilent Technologies, Santa Clara, CA, USA) with the method described by Ortega [23] and 9 minor volatiles by gas chromatography coupled to mass spectrometry (GC-MS; Agilent, Santa Clara, CA, USA) following the method proposed by Yuan and Qian [24].

The extraction of major aroma compounds was performed in dichloromethane using DB-WAX - high-polarity, polyethylene glycol (PEG) column (60 m $\times$ 0.32 mm $\times$ 0.5 µm film thickness) from J&W Scientific (Folsom, CA, USA). For sample preparation, conical bottom glass tubes were used and the following were added: 3.9 g of ammonium sulfate, 6.3 mL of milli-Q grade deionized water 2.7 mL of wine, 20 µL of an internal standard solution (2-Butanol, 4-Methyl-2-pentanol, 4-hydroxy-4-methyl-2-pentanone and 2-Octanol) and 250 µL of dichloromethane. The oven temperature was initially programmed at 40 °C for 5 min, and then ramped to 200 °C. A constant helium flow of 2 mL/min was used. Two mL of aroma extract were injected at 250 °C in splitless mode. Total run time was 75 min per sample. All samples were analyzed in duplicate.

Terpenoids and C13-norisoprenoids (minor volatiles) were determined using headspace-solid phase microextraction coupled with gas chromatography-mass spectrometry (HS-SPME/GC-MS) following the method proposed by Yuan and Qian [24] using an Agilent 6890 gas chromatograph equipped with an Agilent 5973 mass selective detector (Agilent). Two mL of the wine sample were diluted with 8 mL of a citric acid solution (0.5 g/L citric acid, pH 3 saturated with sodium chloride) and 20 µL of 4-octanol (100 µg/L) used as internal standard; all were mixed with a small magnetic stir bar. For volatile extraction a 50/30 µm Divinylbenzene (DVB)/Carboxen (CAR)/Polydimethylsiloxane (PDMS) fiber (Supelco Inc., Bellefonte, PA, USA) was used. The vials for chromatography (20 mL of volume, Agilent Technologies, Santa Clara, CA, USA) were tightly capped and equilibrated at 50 °C in a thermostatic bath for 10 min and extracted by SPME fiber for 50 min at the same temperature with stirring (1000 rpm). To desorb the analytes, the fiber was manually inserted into the injection port of the GC at 230 °C. A DB-Wax column from J&W Scientific (Folsom, CA, USA) (60 m $\times$ 0.32 mm $\times$ 0.5 µm film thickness, Phenomenex, Torrance, CA, USA) was employed to separate the analytes. Carrier gas (helium) was set at a constant flow rate of 1 mL/min. The oven temperature was initially set at 40 °C for 2 min, raised to 230 °C at 5 °C/min for 15 min.

Descriptive sensory analyses were performed by a trained panel of 8 people (4 expert tasters and 4 habitual consumers) from the IMIDRA Institute. This panel had been previously trained in a normalized tasting room in the recognition of wine flavor. Wines were compared by triangle tests (International Organization for Standardization (ISO) 4120:2004) to assess whether aroma differences existed between the different S. cerevisiae fermentations. Sensory descriptive analysis was performed in used to describe and quantify attributes of the wines on the basis of a scale from 1 (low intensity) to 10 (high intensity). The chart included a visual phase (color, color intensity, vivacity/brightness), an olfactory phase (aroma intensity and quality, fruity, vegetal, floral, alcoholic and alteration aroma (off flavour) such as oxidation or microbiologic) and a gustatory phase (alcoholic character, acidity, fruity, vegetal/herbaceous, bitter, body/structure, salty, global taste quality). A hedonic classification was also carried out establishing the order of preference of the samples presented. The final score was obtained as the mean of the wine evaluations with their respective standard deviation and interpreted by graphical representation.

The statistical processing of the data was carried out with software SPSS ver. 20.0 (SPSS, Inc, Chicago, IL, USA). Analysis of variance (ANOVA) was applied on oenological parameters, volatile compounds and sensory attributes of the wines. Tukey honestly significant difference (HSD) post-hoc tests were used to establish the significance of differences between means to assess significance (p 0.05). Principal component analyses (PCA) to evaluate the influence of yeast strains on the volatile composition of wines.

The three studied S. cerevisiae strains could complete the vinifications (residual sugars below 4 g/L) and the time required to carry out the fermentation was the same.

Physico-chemical parameters of wines remained within the legal limits established by European Regulation for table wines elaboration: titratable acidity $\ge$4.5 g/L (Commission Regulation (EC) No 557/94 of 14 March 1994 laying down a transitional measure regarding the total acidity content of the table wine produced in Spain and Portugal and released to the markets in those Member States for 1994), volatile acidity $\le$1.08 g/L (EC 1493/99), pH 2.8‒3.4, alcohol degree $\ge$9–15 (Commission Regulation (EEC) No 2676/90 of 17 September 1990 determining Community methods for the analysis of wines), total SO${}_2$ $\le$210 mg/L (EC 1493/99). All wines obtained presented similar analytical characteristics (Supplementary Table 1) only volatile acidity marked the highest differences. The S. cerevisiae CLI 271 (0.33 g/L) and CLI 889 (0.37 g/L) strains had lower volatile acidity values than commercial control LSA (0.50 g/L). This parameter can play a relevant role in wine aroma and its excessive content is highly detrimental to wine quality. During alcoholic fermentation, the usual quantity of volatile acidity produced by S. cerevisiae is between 0.25 g/L and 0.50 g/L [25], so all Malvasia wines are among this normal range.

The concentration of varietal and major aroma compounds in the wines produced by the tested yeasts was evaluated. Table 1 (Ref. [26, 27, 28, 29, 30, 31]) shows the average and standard deviations of the volatile detected in the different fermentations together with their odor threshold (OTH) [26, 27, 28, 29, 30, 31]. The compounds were classified into chemical families to define their impact on wine and check if there were differences between the yeast strains used. From all the volatile compounds identified, those whose concentrations were higher than their OTH are considered as aroma-contributing compounds. Comparing the total volatiles obtained in the three elaborations, CLI 889 strain wines showed a higher concentration of total volatiles (418.41 mg/L) than those elaborated with CLI 271 and LSA strains (400.78 mg/L and 326.94 mg/L respectively).

The total concentrations of higher alcohols range from 299 mg/L in the wine fermented with LSA to 393 mg/L in the wine fermented with CLI 889 strain. Isoamyl alcohol involved more than 70% of the total alcohols and together with $\beta$-phenylethanol exceeded their OTH in all wines. This compound is related to floral aromas with attributes of roses and is considered to contribute positively to wine aroma [32]. Isobutanol and methionol also contributed to the aroma of wines fermented with CLI 271 and CLI 889 respectively.

The total ester content was higher in wines fermented with CLI 271 strain. Esters are very important for the aroma of wine; they are related to fruity aromas [33]. Isoamyl acetate and ethyl lactate were the dominant esters founded in higher concentrations in wines elaborated with CLI 271 strain. Ethyl isovalerate, ethyl butyrate, ethyl hexanoate, ethyl octanoate and 2-phenylethyl acetate exceeded or were very close to the perception threshold in all wines. These results are in agreement with Balboa-Lagunero [21] in white wines of cv. Palomino fino fermented with different yeast strains where resulting wines from CLI 271 fermentation stood out for their ester content.

As regards fatty acids, butyric, isovaleric, hexanoic and octanoic acids were the main ones found in the wines. The concentration was higher in wines fermented with CLI 271 strain, moreover all of them exceeded the olfactory threshold, therefore they will have a significant aromatic impact on the resulting wines. Although the presence of fatty acids is often associated with off-flavours, they will play an important role in the aromatic balance considering their synthesis antagonistic to the hydrolysis of the ester analogues [20, 32].

The lactone family is characterized by typical fruity aromas. Only $\gamma$-butyrolactone was detected and it resulted in different concentrations for the three preparations. The wine made with LSA showed almost twice the concentration compared to the wine made with strain CLI 271, however, the wine made with strain CLI 889 obtained very similar concentrations to that of LSA. As expected, in none of the cases it was found in concentrations above its perception threshold, this family is mainly found in wood-aged wines to which it contributes important sensory characteristics [34]. Wine elaborated with the LSA strain showed the highest levels of aldehydes and ketones compared to wines fermented with 889 and 271 strains mainly due to the high diacetyl content. This group of compounds is related to oxidation aromas in wine [35]. In previous studies, Palomino must fermented with CLI 271 had the lowest concentration values of some aldehydes that are negatively correlated with wine’s quality; also, these samples obtained the lowest scores in the sensory test for oxidation related descriptors [21].

Regarding the terpene family, the main monoterpenes found were linalool, $\alpha$-terpineol, geraniol and $\beta$-citronellol whose concentrations were variable depending on the yeast strain used, thus, the wines from CLI 889 and LSA strains showed the highest levels of monoterpenes. Linalool, with concentrations above 60 µg/L, was the most abundant and was found above its olfactory perception threshold in all wines. This terpene provides the wine with floral and citrus notes characteristic of Muscat [36]. The compounds analyzed in this family are characterized by their floral and fresh notes. Their presence will be very important because of their influence on wine quality. Traditionally, these compounds have been related to the grapes and not to the fermentation processes, but it has been demonstrated that they are synthesized “de novo” by S. cerevisiae, depending on the redox situation and the level of nitrogen available in the fermentation medium [1]. Therefore, the concentration of free terpenes released into the wine will not only depend on the content of aromatic precursors and the grape variety.

It is important to highlight the significance of $\beta$-damascenone as the only aroma representative of the C13-norisoprenoid family. Its perception threshold is very low and in all three wines exceeds it, providing floral aromas with violet attributes. Similar results were found in Chardonnay wines [37].

In addition, PCA of aroma composition using analytical triplicates of control LSA wines (Fig. 1) was performed in order to reveal the compounds that best differentiated between the Malvasia wines elaborated different yeast strains. According to PCA results, two main components explaining 62.2% of the total variance were retained (with a factor loading $>$0.5000). The first component with 31.4% of the variance explained mainly groups the variables related to wines fermented with LSA strain in the positive plane and those related to wines elaborated with the CLI 889 strain in the negative plane. The main compounds related to factor 1 in LSA wines are $\alpha$-terpinene, diacetyl, ethyl butyrate and ethyl isovalerate. With the exception of $\alpha$-terpinene (whose OTH is not known), all of them were found above the olfactory perception threshold and were therefore relevant in the aroma of these wines. The most significant variables of component 1 related to CLI 889 wines were isoamyl alcohol and cis-3-hexen-1-ol. As for the second component which reveals 30.8% of the variance, it is represented in the positive plane by pleasant aroma compounds ($\beta$-phenylethanol, ethyl hexanoate and isoamyl acetate) but also unpleasant aroma compounds (methionol, benzaldehyde, 1-butanol) where wines from yeast strain CLI 271 are classified, and diethyl succinate and $\beta$-citronellol compounds in the negative plane where wines from yeast strain CLI 889 are located. A clear orderly disposition respect on the aromatic composition of wines have been obtained depending on the S. cerevisiae employed in the fermentation process.

Fig. 1.

Principal component analysis (PCA) of volatile composition data. Malvasia aromatica wines fermented with S. cerevisiae CLI 271, CLI 889 and LSA strains in the squares formed by Principal component 1 (PC1) (31.4%) and Principal component 2 (PC2) (30.8%).

A descriptive tasting of wines was realized at visual, olfactory and gustative levels which results were statistically treated in order to establish the main differences found among the wines fermented with the three yeast strains in study. Differences between wine samples with a significance level of 0.1% (p 0.001) were considered very significant, of 1% (p 0.01) considered significant and of 5% (p 0.05) considered low significant. Visual descriptors were scored very similar between wines, only wine elaborated with LSA strain obtained a score significantly higher than others in terms of vivacity/brightness character (Fig. 2A). In turn, the color of all wines was described as straw yellow with greenish tones.

Fig. 2.

Descriptive analysis of Malvasia wines elaborated with S. cerevisiae CLI 271, CLI 889 and LSA strains. (A)Visual phase. (B) Olfactory phase. (C) Gustative phase. The asterisks * and ** correspond to significant differences between wines from different yeast strains at p 0.05 and p 0.01, respectively.

Regarding aroma descriptors (Fig. 2B), the wines elaborated with CLI 889 and LSA obtained very similar general quality notes (5.6 and 5.8 respectively), very low defects and non-significant differences on olfactory level. Also, these wines reached the best scores in terms of aroma intensity, aroma quality and flowery descriptor. However, CLI 889 wines were considered more alcoholic than others but without significant differences, and together with LSA wines noticeably more floral than wines from CLI 271 fermentations (p 0.01). Tasters highlighted the high floral character of CLI 889 wines with anise notes.

About gustative analysis (Fig. 2C), wines from CLI 889 and LSA fermentations obtained again similar profiles. Both gained the highest outcome for global quality descriptor, with significant differences between CLI 271 and LSA wines (p 0.05). Fruity character was highlighted in CLI 889 wines found significant differences with CLI 271 elaborations (p 0.05). This fruity character of CLI 889 has previously been emphasized in young white wines elaboration [19, 20, 38, 39]. The vegetal/herbaceous and bitter character was superior in CLI 271 wines.

On the other hand, triangular tests were done to determine whether descriptive analysis was determinant to distinguish the samples. Discriminant triangular tastings were carried out using dark glasses and four series of tastings were performed by each type of wine: CLI 889 vs. CLI 271, CLI 889 vs. LSA, CLI 271 vs. LSA. Moreover, the panelists realized a hedonic classification according to their preference. Preferences were taken into account only when tasters answered correctly on each test. After first test (Table 2), the panelists were able to differentiate between wines from CLI 889 and CLI 271 in three of the four series with a 5% of significance level by triangle tests. The tasting panel showed preference by CLI 889 wines in 70% of cases versus 30% of them selected CLI 271 wines. The second test exhibited an elevated percentage of correct responses when CLI 889 and control LSA were compared. In this case, three of four series were distinguished with significance level of 5% (series 2) and of 1% (series 3 and 4), but was not nevertheless manifested preference between the wines (Table 2). Finally, CLI 271 and LSA wines were compared. The level of correct responses was also very significant in this test, the tasters identified clearly the wines from two yeast strains in three of the four series (significance levels of 1% and 5%), and they expressed preference for CLI 271 wine (Table 2).

To estimate the sensory contribution of aromatic compounds to overall aroma of wine, the odor activity value (OAV) was calculated for all aroma compounds in study (Table 3). The OAV is obtained from the ratio between the concentration of certain compound and its perception threshold. Thus, a volatile compound contributes to overall aroma when its concentration is beyond its odor threshold value (OTH); so, odorants with OAV $\ge$1 are considered direct contributors to wine aroma [28, 32]. The wine elaborated with S. cerevisiae CLI 271 strain obtained higher concentrations of $\beta$-phenylethanol (roses), isoamyl acetate, ethyl octanoate and 2-phenylethyl acetate (fruity, fresh). Although other compounds with OAV $>$1 related to unpleasant flavors were found in this wine such as isovalerianic, hexanoic and octanoic acids (cheesy, rancid) (Table 3). This may have contributed to the scores of intensity and global quality aroma were not very high in this CLI 271 wine. On the other hand, there are interesting correlations between volatile compounds and sensory analysis results in wines from CLI 889 and LSA yeast strains. It is worth noting that CLI 889 wine with the highest scores of alcoholic notes (Fig. 2B) also presented the largest concentration of higher alcohols as isobutanol and isoamyl alcohol compounds (OAV $\ge$1) (Table 3). In turn, CLI 889 and LSA strains received high scores in floral character (Fig. 2B), also had more elevated concentration of varietal aromas such as linalool and $\beta$-damascenone. This could mean that both elaborations obtained high intensity and global quality valuations than wines fermented with CLI 271 strain.