*2.4. Statistical Analysis*

Data from chemical analysis of the wines were subjected to one-way ANOVA using STATISTICA 7 (Statsoft, Tulsa, OK, USA) software. Duncan test was carried out to compare mean values and evaluate significant differences. Mean values of volatile compounds were analyzed by principal component analysis (PCA), using JMP® 11 statistical software.

The sensory scores were statistically analyzed and compared according to analysis of variance (ANOVA) using a mixed effect model considering as fixed factors those related to the experimental thesis and as random factors the deviations due to the effect of the judge from the general average of each parameter.

#### **3. Results and Discussion**

Growth kinetics of *S*. *cerevisiae* EC1118 pure culture (control fermentation) and of non-*Saccharomyces*/*S*. *cerevisiae* mixed cultures are reported in Figure 1. In all mixed cultures the initial concentration of the non-*Saccharomyces* yeasts ranged from 10<sup>6</sup> to 10<sup>7</sup> cell/mL. *S*. *cerevisiae* was able to dominate in most of the mixed fermentations and showed, in mixed culture, a growth kinetics that was similar to that of control. In particular, in spite of the presence of the non-*Saccharomyces* yeasts, *S*. *cerevisiae* reached a cell concentration of about 10<sup>8</sup> CFU/mL (T5) and maintained it until the end of the fermentation (T10). When in mixed culture with *H. osmophila*, *S*. *cerevisiae* reached a lower cell concentration (2.5 × 10<sup>7</sup> cell/mL at T5) which remained unvaried for 5 further days of fermentation (T10). Similarly, *L*. *thermotolerans*, even if to a lower extent, affected the growth of *S*. *cerevisiae*, which reached a cell concentration of 4.8 × 10<sup>7</sup> cell/mL and 6.7 × 10<sup>7</sup> cell/mL at T5 and T10, respectively. Conversely, and contrary to that found by Englezos et al. [20], *S*. *bacillaris* did not affect *S*. *cerevisiae* growth. Similar to *H*. *osmophila* and *L*. *thermotolerans, Z*. *florentina* and *T*. *delbrueckii* showed a higher level of competitiveness being still present at concentrations ranging from 1 × 10<sup>4</sup> to 5 × 10<sup>5</sup> cells/mL at T10. On the contrary, *P*. *fermentans*, *S*. *bacillaris*, and *M*. *pulcherrima* persisted at high concentration up to T5 and they almost disappeared at T10.

**Figure 1.** Growth of inoculated non-*Saccharomyces* (-), other non-*Saccharomyces* yeasts (•) and *S. cerevisiae* (). Viable plate counts were done immediately after (T0), and 3, 5, and 10 days after inoculation (T3, T5, and T10, respectively). Data are means ±SD (*n* = 2).

Sugar consumption was consistent with growth kinetics (Figure 2). As expected, in control fermentations (*S*. *cerevisiae* pure culture), it started soon after grape crushing and proceeded faster compared to that observed in all mixed fermentation trials. Among these, the combination *H*. *osmophila*/*S*. *cerevisiae* showed an impairment of sugar consumption with 7.50 g/<sup>L</sup> residual sugar at T10, while the other mixed cultures left from 1 to 1.4 g/<sup>L</sup> residual sugar.

**Figure 2.** Non-*Saccharomyces* and *S. cerevisiae* sugar consumption in each fermentation trial. Different colors indicate different sampling times; T0 (-), T3 (-), T5 (-), T10 (-). Data are means ±SD (*n* = 2).

In most of the mixed cultures ethanol concentrations were comparable to that of the control (Table 2). Mixed starters involving *Z*. *florentina* and *T*. *delbrueckii* were exceptions and produced less ethanol than the control. Lencioni et al. [39], also found slightly lower ethanol concentrations in *Z*. *florentina*/*S*. *cerevisiae* fermentations performed at laboratory scale in white grape must, with respect to the pure *S*. *cerevisiae* culture. Regarding the mixed starter *T*. *delbrueckii*/*S*. *cerevisiae*, decreases in ethanol concentrations, ranging from 0.3% to 0.5%, were also reported by other authors at the end of pilot-scale fermentations [46,47]. Non-*Saccharomyces*/*S*. *cerevisiae* mixed starters have already been proposed as a tool for the possible reduction of ethanol content in wine [19,48–54]. Indeed, the lower ethanol concentration is a consequence of some features of non-*Saccharomyces* yeasts, such as reduced ethanol yield, low fermentation efficiency, and respiro-fermentative metabolism.

**Table 2.** Main analytical parameters of the wines evaluated before bottling.

#### *Fermentation* **2020**, *6*, 63

Mixed starters, including *P*. *fermentans*, *S*. *bacillaris*, and *L*. *thermotolerans*, determined an increase of volatile acidity of about 0.1 g/L, as compared to the control (Table 2). Although non-significant, *H*. *osmophila*/*S*. *cerevisiae* resulted in a more marked increase in volatile acidity which reached values of 0.5 g/L. This is in contrast with previous results obtained with the same yeas<sup>t</sup> strain in co-fermentation with *S*. *cerevisiae*, but at laboratory scale and with a commercial white grape must [38]. Instead, the increase in volatile acidity observed in the fermentation inoculated with *S*. *bacillaris*/*S*. *cerevisiae* is in agreemen<sup>t</sup> with the results obtained by Whitener et al. [55] but in contrast with previously published works showing *C*. *zemplinina* (synonym *S*. *bacillaris*) able to reduce the amount of acetic acid when in mixed culture with *S*. *cerevisiae* [5,10–12]. Nisiotou et al. [56] found lower acetic acid concentration in sequential fermentation *S*. *bacillaris*/*S*. *cerevisiae* carried out as a pilot plant as compared to those performed at laboratory scale fermentation. These discrepancies might be due to the significant strain diversity within this species, as already observed by Englezos et al. [11], but also to a strain specific response to the different fermentation conditions, including the grape variety utilized. In our experimental trials, the presence of other microorganisms, starting from the beginning of the alcoholic fermentations performed at pilot scale, might have interfered with the metabolic activity of the *S*. *bacillaris* yeas<sup>t</sup> strain. Similar observations can be extended to the mixed fermentation conducted with *L*. *thermotolerans* that is usually recognized for low volatile acidity production in wine [57].

The utilization of *P*. *fermentans* and *S*. *bacillaris*resulted also in a significantly higher amount of both total polyphenols and anthocyanins, with respect to the control (Table 2). Recent works indicate that wine color and anthocyanin composition may benefit from the fermentative activity of non-*Saccharomyces* yeasts [58,59]. In particular, it was shown that the inoculation of non-*Saccharomyces*/*Saccharomyces* mixed starters results in higher acetaldehyde production, with effects on anthocyanin-derived pigments [60]. Here, no differences in acetaldehyde content were observed at T10 although its increase at T3 and T5 cannot be excluded in fermentations carried out by mixed starters including *P*. *fermentans* and *S*. *bacillaris*.

Analyses of total polysaccharides, glycerol and volatile compounds, together with the sensory analyses were performed four months after bottling.

Polysaccharides, in particular mannoproteins, impact wine sensorial features by decreasing astringency, improving the mouthfeel and fullness, adding complexity and aromatic persistence, and increasing roundness and sweetness [61–64]. With the exception of those including *H*. *osmophila* and *Z*. *florentina*, all mixed starters produced significantly higher polysaccharides concentrations in respect to the control (Figure 3). In particular, the increase ranged from 2.5% to 33%. In this respect, the most interesting association was *L*. *thermotolerans*/*S*. *cerevisiae* with a final content of total polysaccharides of 732 mg/<sup>L</sup> versus 550 mg/<sup>L</sup> of the control, in agreemen<sup>t</sup> with the result obtained by Gobbi et al. [22] with a different *L*. *thermotolerans* strain. The release of polysaccharides by non-*Saccharomyces* yeasts is not new and a wide intraspecific biodiversity for this characteristic was observed in *Hanseniaspora, Zygosaccharomyces* [4,35,37,38] and *Schizosaccharomyces* yeasts [34].

The concentration of glycerol, responsible for the sweetness of red and white wines [65,66], was significantly higher in most of the wines produced by mixed starters, apart from those including *H*. *osmophila, P*. *fermentans* and *Z*. *florentina* (Figure 3). As expected, the association *S*. *bacillaris*/*S*. *cerevisiae* resulted in the highest glycerol concentration (11.4 g/L), in accordance with that already observed for the species *S*. *bacillaris* [25,67,68].

**Figure 3.** Total polysaccharides (-) and glycerol (-) in wines obtained four months after bottling. Data are means ±SD (*n* = 2). Values displaying different letters (a, b, c, d) are significantly different according to the Duncan test (*p* ≤ 0.05).

The concentrations of the main volatile compounds are reported in Table 3. The concentrations of acetaldehyde, propanol, and hexanol produced by mixed starter cultures were comparable to that of the control. In contrast, mixed starters resulted in higher production of some of the higher alcohols. In particular, significant increases of 2-methyl-1-propanol (isobutanol) were observed in respect to control fermentation (47 ± 2 mg/L). This compound ranged from a minimum of 66 mg/<sup>L</sup> (for the associations including *H*. *osmophila* and *Z*. *florentina*) to a maximum of 123 ± 8 mg/<sup>L</sup> in the wine produced by *S*. *bacillaris*/*S*. *cerevisiae*. However, the sum of amylic alcohols (i.e., 2-methyl-1-butanol, 3-methyl-1-butanol) was significantly higher in wines produced with the mixed starters including *M*. *pulcherrima* (328 mg/L), *T*. *delbrueckii* (299 mg/L) and *L. thermotolerans* (294 mg/L) than in the control wine (266 mg/L). In agreemen<sup>t</sup> with that reported by other authors [69–72], mixed starters including *Hanseniaspora, Pichia* and *Zygosaccharomyces* showed lower production of higher alcohols. Interestingly, all the wines obtained with mixed starters presented significantly higher concentrations of 2-phenylethanol (8–9.5 mg/L) (which provides a rose-like flavor) compared to the control (6.1 mg/L). In particular, the highest concentrations of 2-phenylethanol were reached in mixed fermentations including *M*. *pulcherrima* (9.2 mg/L) and *T*. *delbrueckii* (9.4 mg/L). These results agree with those found by other authors showing that *M. pulcherrima* and *T*. *delbrueckii* produce high level of 2-phenylethanol [3,73,74]. Ethyl acetate was the main ester produced. At high concentrations (>100–150 mg/L) ethyl acetate determines a solvent-like aroma. Interestingly, with the exception of the associations including *M*. *pulcherrima* and *H*. *osmophila* that nearly doubled the amount produced by *S*. *cerevisiae* starter culture (54 mg/<sup>L</sup> and 51 mg/L, respectively), the other mixed starters determined slight increases in ethyl acetate in respect to the control. In any case, ethyl acetate concentration was always below the perception threshold (Table 3). These findings confirm those already observed in other studies [37,38], where some non-*Saccharomyces* yeasts, generally considered spoilage yeasts, produced in mixed culture ethyl acetate concentrations below those normally produced by the relevant pure culture. On the other hand, many studies report that most non-*Saccharomyces* yeasts can produce high amounts of ethyl acetate [71,75]. However, this discrepancy may be due to the wide inter-generic and intra-generic variability observed for the production of this ester compound. Accordingly, Domizio et al. [38] found, by analyzing eleven yeas<sup>t</sup> strains of *Hanseniaspora* (belonging to four different species), that ethyl-acetate production ranged from 27 to 333 mg/L. It is also worth underlining that this compound, at low concentration, might contribute to wine fruity aroma.


**Table 3.** Volatile compounds (mg/L) of wines four months after bottling.

#### *Fermentation* **2020**, *6*, 63

Among other acetates analyzed, 2-phenylethyl acetate (with a fruity and flowery flavor) was significantly higher only in wines fermented by *H. osmophila*/*S*. *cerevisiae* (0.016 mg/L) in comparison with the control wine (0.003 mg/L). This result is in agreemen<sup>t</sup> with the known capacity of this yeas<sup>t</sup> species to release high levels of 2-phenylethyl acetate [70–72,76].

Other ethyl esters compounds, such as ethyl lactate, ethyl butyrate, ethyl hexanoate and ethyl octanoate, were present in all the wines with similar or slightly lower concentrations in comparison to those present in the control wine. An exception, regarding ethyl lactate, was made for wines produced by the association of *L. thermotolerans*/*S*. *cerevisiae* that reached a concentration about 3-fold higher than that measured in the control wine (4.7 mg/L). This result is in accordance with that observed in previous studies [22,58,77], and is compatible with lactic acid production by *L. thermotholerans* [78].

PCA analysis showed evident differences among the strains tested as a function of volatile compounds production and this reflects the ability of each strain to give a specific aromatic imprint to the final wines (Figure 4). Based on volatile compounds content in the resulting wines, *H. osmophila* and *M. pulcherrima* were positioned in the upper left quadrant and characterized by acetate esters and 2-phenyl ethanol. *S*. *bacillaris* and *P. fermentans* were placed in bottom left quadrant characterized by the production of isobutanol. *T. delbrueckii* and *L. thermotolerans* were placed in the upper right quadrant due to the production of isoamyl alcohol and isoamyl acetate, while *Z. florentina* and *S*. *cerevisiae* control strain were positioned in the right bottom quadrant and were characterized by the production of ethyl esters.

**Figure 4.** Principal component analysis (PCA) based on the production of volatile compounds.

According to the results of sensory analysis carried out four months after bottling, all the wines obtained with mixed fermentation starters were perceived as significantly more provided in color intensity, in respect to the control wine. This was particularly true for wines obtained with associations including *S*. *bacillaris* and *M. pulcherrima* (Figure 5). This result is in accordance with the higher amounts of total polyphenols and anthocyanins found in the relevant wines, and in respect to the control. Moreover, it agrees with the findings of other authors pointing out that many non-*Saccharomyces* yeasts in sequential fermentation with *S*. *cerevisiae* may enhance color intensity of wines, promoting the formation of derivatives with more stable color than anthocyanins [79–82]. This is particularly important for Sangiovese wine that, being rich in unstable and oxidizable phenols, is characterized by limited color stability [83] and suggests that the utilization of mixed starters, including non-*Saccharomyces* yeasts, might represent an option for the managemen<sup>t</sup> of Sangiovese color stability.

**Figure 5.** Sensory perception of color in wines 4 months after bottling (QDA score: scale 1–4). Values displaying different letters (a, b, c, d) are significantly different according to the Duncan test (*p* ≤ 0.001).

Despite the differences found in the relevant volatile compounds profile, no significant differences among the wines were found in the aromatic profile (descriptors: floral, fruity, canned fruits, spicy, candy, chemical, earthy).

Concerning the taste descriptors used in the organoleptic assessment of wines, significant differences resulted only regarding astringency (*p* ≤ 0.01) and mouth dryness (*p* ≤ 0.001) perception (Figure 6). In particular, while the association including *P. fermentans* resulted in a more astringent wine, in respect to the control, that including *T. delbrueckii* emerged as less astringent with respect to the control wine and all the other wines.

**Figure 6.** Sensory perception of astringency (-) and dryness (-) in wines 4 months after bottling (QDA: scale 1–4) Values displaying different letters (a, b, c, d) are significantly different according to the Duncan test (for astringency at *p* ≤ 0.01 and for dryness at *p* ≤ 0.001).

The perception of mouth dryness was higher in wine deriving from the mixed starter including *M*. *pulcherrima*, while the association *T*. *delbrueckii*/*S*. *cerevisiae* proved the most effective in reducing this sensation in the mouth.
