Effect of a Multistarter Yeast Inoculum on Ethanol Reduction and Population Dynamics in Wine Fermentation

Microbiological strategies are currently being considered as methods for reducing the ethanol content of wine. Fermentations started with a multistarter of three non-Saccharomyces yeasts (Metschnikowia pulcherrima (Mp), Torulaspora delbrueckii (Td) and Zygosaccharomyces bailii (Zb)) at different inoculum concentrations. S. cerevisiae (Sc) was inoculated into fermentations at 0 h (coinoculation), 48 h or 72 h (sequential fermentations). The microbial populations were analyzed by a culture-dependent approach (Wallerstein Laboratory Nutrient (WLN) culture medium) and a culture-independent method (PMA-qPCR). The results showed that among these three non-Saccharomyces yeasts, Td became the dominant non-Saccharomyces yeast in all fermentations, and Mp was the minority yeast. Sc was able to grow in all fermentations where it was involved, being the dominant yeast at the end of fermentation. We obtained a significant ethanol reduction of 0.48 to 0.77% (v/v) in sequential fermentations, with increased concentrations of lactic and acetic acids. The highest reduction was achieved when the inoculum concentration of non-Saccharomyces yeast was 10 times higher (107 cells/mL) than that of S. cerevisiae. However, this reduction was lower than that obtained when these strains were used as single non-Saccharomyces species in the starter, indicating that interactions between them affected their performance. Therefore, more combinations of yeast species should be tested to achieve greater ethanol reductions.


Introduction
In recent years, the average ethanol concentration in wine has increased, mainly due to climate change and consumer preference for wine styles [1][2][3]. Different strategies have been applied to reduce ethanol production in wine, such as decreasing the leaf area to lower the sugar content in grape berries, reducing the maturity of grapes, removing sugar from the grape must, developing or screening low-alcohol wine yeasts and removing alcohol from wine (reviewed by [4]). Compared with microbiological strategies, other strategies might have negative effects on wine, such as delaying maturity, reducing the yield of berries, causing a significant reduction in anthocyanins, soluble solids and volatile compounds, and decreasing the wine color [5][6][7][8]. Therefore, microbiological strategies were considered to be effective and accompanied by smaller impacts on the wine sensory profile and quality. In particular, the use of non-Saccharomyces yeast strains for reducing the alcohol content of wines has been proven to improve the wine aroma complexity and has become a consistent proposal [9][10][11].

Fermentation Trials in Natural Must
Eight different mixed fermentations were carried out in this study (Table 1). Initially, all fermentations were simultaneously inoculated with three non-Saccharomyces strains at two different concentrations: 10 6 cells/mL (fermentations named 1CN) or 10 7 cells/mL (fermentations named 10CN). Additionally, Sc was inoculated into these two mixed non-Saccharomyces starters at the same moment (coinoculated fermentations were named 1CA and 10CA) and at 48 or 72 h later (sequential fermentations named 1S48, 1S72, 10S48 and 10S72, respectively), always at the same concentration (10 6 cells/mL). A single fermentation with Sc (10 6 cells/mL) was used as the control fermentation (C).
Triplicate fermentations were conducted in 250 mL glass bottles with 230 mL of NM and incubated at 22 • C with stirring at 120 rpm. The bottle cap had two ports-one connected to a 0.22 µm filter (Dominique Dutscher, Brumath, France) for gas flow and the other clamped by an iron clip for sampling. Before inoculation with Sc in sequential fermentations, the concentration of assimilable nitrogen in the fermented must was determined and supplemented, if needed, to 100 mg/L. Fermentations were monitored by measuring the must density with an electronic densitometer (Densito 30PX Portable Density Meter, Mettler Toledo, Hospitalet de Llobregat, Spain). Fermentations were considered finished when the residual sugars were below 2 g/L, checked by D-glucose/D-fructose enzymatic assays (Biosystems S.A., Barcelona, Spain) in the Miura autoanalyzer, or arrested when the must density did not decrease for more than two days. Samples from the end of fermentation were centrifuged at 7800 rpm for 5 min, and supernatants were frozen at 20 • C until chemical analysis.

Colony Counting
The yeast population during fermentation was analyzed by colony growth on WLN plates according to the different colony morphologies of Saccharomyces and non-Saccharomyces strains [35]. Briefly, samples were serially diluted in sterilized Milli-Q water from a Milli-Q purification system (Millipore S.A.S., Molsheim, France). The number of colony-forming units per milliliter (CFU/mL) was determined by plating 100 µL of three appropriately chosen dilutions on WLN agar. The plates were incubated at 28 • C for 2 or 3 days.

PMAxx Treatment
To obtain only the DNA of living cells, PMAxx TM viability dye (Biotium Inc., Fremont, CA, USA) was added to must samples as previously described by Navarro et al. [38]. Briefly, selected samples of must (1 mL) were centrifuged, and the pellets were washed with sterilized distilled water and treated with 25 µM PMAxx. After incubation for 10 min in darkness, PMAxx was permanently linked to the DNA of dead cells by subjecting samples twice to light for 30 s, with an interval of 1 min on ice. Pellets were recovered by centrifugation and frozen until DNA extraction.

DNA Extraction and qPCR Analysis
DNA was extracted using the DNeasy Plant Mini Kit (QIAGEN GmbH, Hilden, Germany) following the instructions of the manufacturer. Cell quantification of the different yeast species was conducted by species-specific qPCR using the primers CESP-F/SCER-R for S. cerevisiae [31], MP2-F/MP2-R for M. pulcherrima [41], Tods L2/Tods R2 for T. delbrueckii [42] and ZBF1/ZBR1 for Z. bailii [29] (all primer sequences are shown in Table A1). For all samples, qPCR was performed in triplicate using TB Green TM Premix Ex Taq TM II (2×) (Takara Bio Inc., Kusatsu, Japan) with a final volume of 20 µL (2 µL of DNA, 0.8 µL of each primer, 0.08 µL of ROX Reference Dye (50×), 10 µL TB Green Premix Ex Taq II (2×) and 6.32 µL of sterilized Milli-Q water) on a QuantStudio TM 5 real-time PCR instrument (Applied Biosystems by Thermo Fisher Scientific, Waltham, MA, USA). A 96-well nonskirted PCR plate (4titude ® Ltd., Wotton, UK) was used for the reaction. The amplification reaction was one cycle of 95 • C for 1 min and 40 cycles of 95 • C for 5 s and 60 • C for 35 s, followed by a dissociation step. Milli-Q water was used as a negative control. Ct (the cycle threshold) was determined using Thermo Fisher Scientific software (Waltham, MA, USA).
Standard curves were calculated for each species with and without PMAxx treatment, as described by Navarro et al. [38] but slightly modified. After incubating the yeast colony in YPD liquid medium at 28 • C for 24 h, samples with 10 8 cells/mL were collected in triplicate and centrifuged at 10,000 rpm for 2 min. The pellet was washed once with 1 mL sterilized distilled water and was subjected to PMAxx treatment as previously described. The pellet was stored at −20 • C until DNA extraction. Standard curves were created by plotting the average Ct values of a tenfold serial dilution of DNA from 10 8 to 10 cells/mL against the log of cells/mL.

Chemical Analysis
Residual sugars of samples at the end of fermentation were quantified by D-glucose/ D-fructose enzymatic assays. Ethanol, glycerol and organic acids (citric acid, malic acid, tartaric acid, acetic acid, lactic acid and succinic acid) in the samples were determined by high-performance liquid chromatography (HPLC) using an Agilent 1100 (Agilent Technologies, Waldbronn, Germany) as previously described by Quirós et al. [43] and Zhu et al. [39].

Statistical Analysis
All graphs were generated using GraphPad Prism ® version 8 (GraphPad Software, San Diego, CA, USA). The results are expressed as the mean ± standard deviation (SD). Statistically significant differences (one-way ANOVA) were analyzed by IBM SPSS Statistics version 23.0 (IBM, NY, USA). The ethanol yield was calculated with the Equation (1).

Fermentation Kinetics
The fermentation kinetic profiles obtained under the different inoculation conditions, measured as must density reduction, are shown in Figure 1. The fastest fermentations were those with single inoculation with Sc and with the coinoculation of all the strains at the same time (1CA and 10CA), which showed similar fermentation profiles and completed fermentations in 7-8 days (10CA completed the fermentation 1 day earlier than that of Sc alone (C)). The fermentation kinetics were less affected by the inoculum concentration of non-Saccharomyces yeasts but were influenced by the inoculation time of Sc. The sequential inoculations, in which only non-Saccharomyces strains were fermented for 48-72 h, resulted in slower fermentations, finishing all the sugars in 11 (1S48, 10S48, 10S72) or 13 days (1S72) (Figure 1). These fermentations were slightly slower when Sc was inoculated later, at 72 h, mainly in the lower non-Saccharomyces inoculum (1S72). Finally, fermentations with only non-Saccharomyces species (1CN and 10CN) showed the slowest fermentation kinetics, getting stuck on the 13th day, at densities of 1.002 and 1.001 g/mL, respectively (Figure 1a,b). In general, the fermentations with non-Saccharomyces strains inoculated at 10 7 cells/mL showed slightly faster fermentation kinetics than did those inoculated at 10 6 cells/mL (Figure 1a Table 1). The initial non-Saccharomyces inoculum size was 10 6 cells/mL (a) and 10 7 cells/mL (b).

Yeast Population Dynamics
To study the yeast population dynamics during single or mixed fermentations, the growth of different species was obtained by different methodologies (Figure 2). On one hand, viability was determined during the fermentation process by plating the samples on WLN agar. Due to the different colony morphologies of the four yeast species on WLN agar [35], the colonies of these four strains could be counted on the same plate. On the other hand, PMA-qPCR analysis was performed to quantify the living yeast population using specific primers for each species. We applied PMA-qPCR from the second day of fermentation because the adaptation of the cells to the medium may produce an underestimation of the population by qPCR [38].
Yeast population analysis based on Wallerstein Laboratory Nutrient (WLN) plates (solid lines ) and qPCR with PMAxx treatment (dotted lines ), determined in the different fermentative conditions: non-Saccharomyces multistarter coinoculated with (a,b) and without (g,h) S. cerevisaie (Sc); non-Saccharomyces multistarter with sequential inoculation of Sc at 48 h (c,d) or 72 h (e,f) and the single Sc fermentations (i). The initial non-Saccharomyces inoculum size was 10 6 cells/mL (a,c,e,g) and 10 7 cells/mL (b,d,f,h). Each species is shown in different colors (Mp in red, Td in green, Zb in blue and Sc in black).
First, the standard curves were calculated by plotting the Ct values (<30) of DNA from 10 3 to 10 8 cells/mL (M. pulcherrima was from 10 4 to 10 8 cells/mL) against the log input cells/mL (Table A2), with efficiencies between 87.92% and 98.83%. First, the standard curves were calculated by plotting the Ct values (<30) of DNA from 10 3 to 10 8 cells/mL (M. pulcherrima was from 10 4 to 10 8 cells/mL) against the log input cells/mL (Table A2), with efficiencies between 87.92% and 98.83%.
After yeast inoculation, all strains were able to grow in fermentation media and showed different population dynamics, as seen in the results obtained from WLN plates. In fermentations in which each non-Saccharomyces species was inoculated at 10 6 cells/mL, the population of non-Saccharomyces yeasts grew rapidly on the first day (Figure 2a,c,e,g). Td quickly increased the population, reaching up to 2 × 10 8 CFU/mL on the third day. The Td population was then stable until the end of fermentation, becoming the dominant yeast in all mixed inoculations, except in the coinoculated fermentations with Sc (1CA and 10CA, Figure 2a,b), in which Sc significantly impaired growth, with the fermentation reaching a maximum Td population of only 10 7 CFU/mL under 1CA conditions. Mp reached a maximum population of 2.7 × 10 7 CFU/mL in the sequential or non-Saccharomyces coinoculations, with densities higher than that of Zb within the first days but decreasing after the sixth day under most conditions (the Mp decrease occurred much earlier in the coinoculation with Sc, 1CA). In the case of Zb, the maximum growth was reached in the late fermentation stage, with a population of 2.8 × 10 7 CFU/mL, and was maintained rather stably until the end of fermentation.
The population dynamics in fermentations inoculated with 10 7 cells/mL of each non-Saccharomyces yeast differed from those inoculated at 10 6 cells/mL (Figure 2b,d,f,h).
In those fermentations, Td and Zb presented similar population dynamics, and although Td was again the yeast with the highest population among all non-Saccharomyces yeasts, the difference relative to Zb was not as large (Td reached 9.7 × 10 7 CFU/mL and Zb 7.0 × 10 7 CFU/mL). Mp achieved the highest population (5.0 × 10 7 CFU/mL) on the second day and became the minority strain until the end of fermentation.
Regarding the S. cerevisiae dynamics, the dominance was only achieved in both coinoculated fermentations (1CA and 10CA), reaching a similar concentration to that of the Sc single inoculation on the third day of fermentation (1.8 × 10 8 CFU/mL, Figure 2a,b). However, the growth of Sc was restricted in sequential fermentations, where its population increased gradually until the end of fermentation, achieving up to 4.7 × 10 7 CFU/mL (Figure 2c-f). In these sequential fermentations, Sc together with Td were the majority strains at the end of the fermentation. This loss of the full imposition was not due to the lack of assimilable nitrogen, since supplementation to 100 mg/L was carried out in the deficient fermentations (1S72, 10S48 and 10S72) before the inoculation of Sc (as detailed in Table A3).
As already mentioned, during the fermentation process, Td and Zb colonies were observed until the end of fermentation (except in fermentation 1CA, where Td and Zb were absent at the end point), while Mp was absent on WLN agar after mid-fermentation. However, PMA-qPCR analysis allowed the detection of Mp until the end of fermentation, with the population ranging from 1 × 10 5 to 1.1 × 10 6 cells/mL, as well as that of Td and Zb at the end of fermentation in 1CA (Figure 2a). After the second day of fermentation, the trends of the yeast populations of Td, Zb and Sc obtained by PMA-qPCR analysis were similar to those of the WLN counting, although the cell concentrations obtained by PMA-qPCR were higher, reaching a difference of one order of magnitude for most non-Saccharomyces yeasts (Figure 2). Specifically, for the Td, Zb and Sc strains, the maximum populations obtained by PMA-qPCR from mixed fermentations reached 7.6 × 10 8 , 1.7 × 10 8 and 1.2 × 10 8 cells/mL, respectively. In contrast, the population of Mp obtained by PMA-qPCR analysis before mid-fermentations was lower than that obtained by WLN counting in the first 6 days for fermentations inoculated with 10 6 cells/mL and in the first 3 days for fermentations inoculated with 10 7 cells/mL.

Main Fermentation Byproducts
To detect the ethanol reduction of the different inoculated fermentations and the metabolic characteristics of yeasts under the different conditions, the residual sugar, ethanol production and main byproducts at the end of fermentation were analyzed, as shown in Table 2.
All fermentations with a mixed inoculum of non-Saccharomyces and Sc were completed, with less than 2 g/L residual sugars. However, the fermentations with only non-Saccharomyces strains, 1CN and 10CN, presented higher concentrations of residual sugars due to the stagnation of these two conditions ( Table 2). Among all fermentations, the control fermentation (C) had the highest ethanol production (13.06%, v/v) and ethanol yield (0.47 g/g) (Figure 3). Mixed fermentations decreased ethanol production by 0.13 to 0.77% (v/v) compared to single fermentation by Sc (C), and this decrease was significant in 5 fermentations (1S48, 1S72, 10CA, 10S48 and 10S72) ( Table 2). Fermentation 10S72 showed the highest ethanol reduction with the lowest ethanol yield (0.44 g/g), followed by 10S48, 1S72, 1S48 and 10CA (Table 2, Figure 3). Among these sequential fermentations, our results show a trend in relation to the inoculum concentration of non-Saccharomyces yeast, since a higher ethanol reduction was achieved when a higher inoculum of non-Saccharomyces was used (0.65 ± 0.11 and 0.77 ± 0.06 with 10 7 cells/mL vs. 0.48 ± 0.10 and 0.49 ± 0.08 with 10 6 cells/mL). Values are mean ± standard deviation of three independent replicates; The initial sugar concentration of synthetic must was 220 g/L; * means statistically significant differences from the control sample of C on the same row (LSD test, p < 0.05); NC means no calculate ethanol reduction due to stuck fermentation. Regarding the other fermentation byproducts, the concentration of glycerol in the mixed fermentations was similar to that of single fermentation with Sc (C), except in the 1S48, 1CN and 10CN fermentations, which had lower glycerol contents, although the latter two conditions had residual sugars left. On the other hand, in all fermentations, the concentrations of lactic and acetic acids were higher in the mixed fermentations, although they remained below 0.33 g/L and 0.36 g/L, respectively (Table 2). Surprisingly, more acetic acid was detected at a lower inoculum concentration of non-Saccharomyces yeasts (10 6 cells/mL).

Discussion
In recent years, non-Saccharomyces yeasts have been proposed for use as starters in alcoholic fermentation to reduce the ethanol content in wine [13,15,18,22,39,[44][45][46]. Most studies have focused on evaluating the effects of single or mixed starters composed of one S. cerevisiae strain and one non-Saccharomyces species. Previous studies showed that as single starters, strains of M. pulcherrima and Z. bailii were not able to complete fermentation or had a low fermentation capacity [21,44], while T. delbrueckii has been reported to be a strong fermenter and able to finish fermentations [22,47]. However, fermentations inoculated with several non-Saccharomyces species as a multistarter have been poorly studied. In the current study, we evaluated the use of a multistarter of three non-Saccharomyces strains in Muscat grape fermentation. These strains were selected in a previous study [39], in which several strains of these three species and other non-Saccharomyces species were screened for their ability to reduce ethanol. When this multistarter of non-Saccharomyces strains was used, we observed stagnation of fermentation, even though Td was the dominant strain throughout the fermentation process and has shown its ability to complete fermentation when used as a single inoculum [47][48][49]. This lower performance of Td when used in multistarter fermentation could be due to interspecific microbial interactions, such as competition for nutrients [50][51][52], or the production of inhibitory compounds, such as killer toxins or other antimicrobial peptides or vesicles [53][54][55][56], as part of cell-cell interaction mechanisms. The higher population size of the non-Saccharomyces inoculum sped the consumption of sugars until mid-fermentation, although this was not enough to complete the fermentation, and similar final populations and residual sugars were obtained under both conditions. Mp was the minority strain under all conditions, and its population decreased significantly throughout the fermentation process, which agrees with previous studies [26,57,58].
As most non-Saccharomyces yeasts are incapable of completing alcoholic fermentation, S. cerevisiae is usually added, either simultaneously as a coinoculum or sequentially at 24-72 h after non-Saccharomyces yeast inoculation. These inoculation strategies reduce the risk of a stuck fermentation [59,60]. Indeed, in the current study, only the fermentations involving S. cerevisiae were complete. When all strains were coinoculated, Sc was the dominant strain and became the most abundant yeast at the end of fermentation, regardless of the inoculum concentration of non-Saccharomyces yeasts. The dominance of Sc during the fermentation process could be explained by its high ability to tolerate different stresses, such as high ethanol levels, especially relevant in the last stage of fermentation [61]. Additionally, different mechanisms, such as cell-to-cell contact, nutrient competition, secretion of toxic compounds or changes in media [26,[62][63][64] could also be responsible of the dominance. The effect of metabolite production and changes in the fermentative medium produced by S. cerevisiae has been proven to reduce the competition and persistence of M. pulcherrima [65], which agrees with the drastic reduction in the Mp population observed in fermentations when Sc was coinoculated. As described above, the fermentation kinetics were mainly affected by the inoculation time of Sc. Specifically, coinoculated fermentations were faster and finished earlier compared to sequential fermentations, which agrees with previous studies [64,66].
Focusing on competition for nutrients, a key factor in wine fermentation is the availability of assimilable nitrogen. On one hand, in pure-culture fermentations, some non-Saccharomyces yeasts (H. uvarum, H. vineae, S. bacillaris and T. delbrueckii) consume less assimilable nitrogen than does S. cerevisiae [47,67,68]. On the other hand, studies have shown that in sequential fermentations, some non-Saccharomyces species, such as T. delbrueckii, L. thermotolerans and S. bacillaris, are capable of consuming almost all assimilable nitrogen within 48 or 72 h before S. cerevisiae is inoculated, which could result in incomplete fermentation due to low growth of S. cerevisiae [47,69]. In the current study, to avoid stuck fermentations due to nitrogen limitation and to promote the growth of Sc, assimilable nitrogen was restored before Sc inoculation, as non-Saccharomyces yeasts had been consuming up to 80% of assimilable nitrogen present in the must. As expected, this addition produced a quick increase in the population of Sc. This conclusion is supported by Medina et al. [68], who also observed an increase in the percentage of the Saccharomyces strain after assimilable nitrogen addition.
A critical aspect for studying yeast interactions during fermentation is to evaluate the yeast population, especially living cells. The traditional method used for determining the yeast living cells during alcoholic fermentation is based on colony counting in different culture media, which takes 2 or 3 days and might cause some deviations [27,70]. In mixed or spontaneous fermentations, to differentiate and count the different yeast species, we need to use selective or differential media [27,41]. One of these media is WLN agar, which is able to distinguish different yeast species according to their colony morphology [25,38,71]. In our work, the four yeast species had different colony morphologies on this medium, which allowed their differential counting. In comparison, qPCR analysis is a relatively fast and sensitive technique to simultaneously identify and quantify different targeted yeast species [29,41,72]. However, as described above, this method cannot distinguish the populations of viable and dead cells some DNA-binding dyes, such as PMA, are used and combined with qPCR, allowing the quantification of only the viable cell population [34,38,73].
In the current study, cell counting on WLN plates and PMA-qPCR were used to monitor the viable populations of the different yeast species during the fermentation process. Specific primers for qPCR were available for all species used in the current work [34,38], which presented a high efficiency and good quantification limit. As described above, even though yeast population profiles were similar between WLN counting and PMA-qPCR analysis, they still showed some differences. First, in the coinoculated fermentations with all strains (1CA and 10CA), Sc was the dominant species at the end of fermentation, without detection of these species in the WLN counting in 1CA. Indeed, WLN medium has been described to be useful for quantifying species with similar log populations but is unable to detect species with a population 1 or 2 logs lower than the main species [38], which would explain the inability to detect some species, especially Mp, when its population started decreasing. In contrast, PMA-qPCR sensitively detected the populations of all non-Saccharomyces strains until the end of fermentation, even at low concentrations, as observed in previous studies [32,38,41,74]. Surprisingly, the viable cell population of Mp obtained from PMA-qPCR analysis was lower than that from WLN counting before mid-fermentation. This could be explained by the different membrane compositions and permeabilities of this species. Vázquez et al. [75] showed that the cellular lipid composition of M. pulcherrima during grape must fermentation contains a high percentage of polyunsaturated fatty acids, which results in more fluid membranes. Moreover, the permeability of the cell membrane is known to be increased upon contact with grape must or ethanol [76]. Thus, this higher permeability could facilitate the penetration of PMAxx into viable Mp cells, resulting in lower quantification by PMA-qPCR analysis. Navarro et al. [38] described this effect in different strains in the early stage of fermentation.
In recent years, M. pulcherrima, T. delbrueckii and Z. bailii have been proven to be species able to reduce ethanol content, mainly in mixed fermentations with S. cerevisiae. For example, M. pulcherrima was able to reduce 0.7-1.5% (v/v) ethanol in sequential fermentations in grape must or defined medium [11,15,39,44], while T. delbrueckii and Z. bailii achieved ethanol reductions of 0.6-1.0% (v/v) and 0.7-1.8% (v/v), respectively, in sequential fermentations with S. cerevisiae compared to single S. cerevisiae fermentation [3,11,39]. All these ethanol reductions were achieved with fermentations initiated by a single non-Saccharomyces strain; however, this reduction changed when a multistarter inoculation was used. For example, according to the results from Varela et al. [23], in sequential fermentations with M. pulcherrima or S. uvarum, the ethanol content decreased by 1.09 and 1.01% (v/v), respectively, while when those species were simultaneously inoculated as mixed starters with S. cerevisiae, the ethanol content decreased by 1.85% (v/v), showing an additive effect of coinoculation. In the current study, we used a mixed starter of three non-Saccharomyces species selected based on their ability to reduce ethanol. Our previous studies showed that the selected Mp, Td and Zb strains were able to reduce the ethanol content by 1.39, 0.84 and 1.02% (v/v), respectively, in sequential fermentations with Sc (inoculated at 48 h), using the same natural must [39]. In this research, the mixed inoculation of these three species in sequential fermentations with Sc resulted in a lower ethanol reduction (less than 0.77% (v/v)) compared to sequential fermentation with a single non-Saccharomyces strain. Thus, we did not observe an additive effect, as did Contreras et al. [77] or Varela et al. [23]. Instead, those species showed a dissipative impact on reducing ethanol production when used together. Similar results were also observed in the study of Contreras et al. [77], in which sequential fermentations by the single starter of M. pulcherrima achieved an ethanol reduction of 1.16-1.76% (v/v), while sequential fermentations initiated by mixed starters of M. pulcherrima, H. uvarum, P. kluyveri and T. delbrueckii reduced the ethanol content by only 0.38% (v/v). Therefore, Contreras et al. [77] demonstrated that ethanol reduction is very dependent on the yeast combination used in the starter, since a yeast strain with a high ability to reduce ethanol, such as M. pulcherrima, was affected by the presence of other yeasts, and as a result, its ability to reduce ethanol could be weakened or strengthened. Moreover, the relative proportion between yeast species during the fermentation could affect the metabolite production. Indeed, comparisons between pure cultures and sequentially inoculated cultures revealed changes in the distribution of carbon fluxes during fermentation [78]. Finally, the initial must composition and nutrient availability can also have an impact in the metabolite production, as previously demonstrated by Seguinot et al. [78].
In addition, our results showed that the inoculum size of non-Saccharomyces yeasts also has an impact on ethanol production, with a lower ethanol content in wines inoculated with higher populations of non-Saccharomyces yeasts. This agrees with the results obtained by Maturano et al. [79], in which a higher inoculum size of non-Saccharomyces yeasts (5 × 10 6 cells/mL vs. 1 × 10 6 cells/mL) produced wines with a lower ethanol content. Thus, more combinations of yeast species in terms of both the number and diversity of strains and the amount of each strain inoculated should be tested to find the best combination for ethanol reduction.
Non-Saccharomyces yeasts usually differ from S. cerevisiae in the distribution of their metabolic flux during fermentation [46,67,78,80]. Indeed, several non-Saccharomyces species are able to aerobically respire sugar, which results in altered formation of the main metabolites produced during fermentation, including ethanol, glycerol and organic acids [3,20,45,46]. However, although ethanol production can be decreased when providing non-Saccharomyces yeasts with oxygen, this sometimes has undesirable side effects, such as higher acetic acid or ethyl acetate levels [3,81]. In this study, a significant decrease in ethanol levels was observed in mixed fermentations, even without the addition of oxygen and without a detrimental increase in acetic acid. Indeed, higher levels of acetic and lactic acids were obtained in the presence of non-Saccharomyces yeast, mainly in sequential inoculations, but the levels were kept below 0.33 g/L for lactic acid and 0.36 g/L for acetic acid, with nondetrimental concentrations for the final wines. On the other hand, the concentration of glycerol did not increase in our reduced-ethanol wines, which implied that the production of glycerol was not the main pathway of ethanol reduction in mixed culture fermentations, as occurred in other studies [39,82]. Nevertheless, the final amounts of acetic and lactic acids are not enough to counterbalance the decrease in ethanol and glycerol. Previous studies had also observed that the reduction of the ethanol yield by some non-Saccharomyces strains, could not be fully explained by the overproduction of glycerol or organic acids, suggesting that respiration would be responsible, at least in part, of the poor ethanol yield observed for these strains [83,84]. However, the analysis of the volatile composition of wines (higher alcohols, volatile fatty acids, esters, carbonyl compounds) would help the understanding of the effect of this type of inoculation on the overall complexity of wines [11,85].
In summary, our results confirm that PMA-qPCR analysis is a fast and sensitive method for monitoring the viable cell population dynamics in mixed fermentations of T. delbrueckii, M. pulcherrima, Z. bailii and S. cerevisiae. T. delbruecckii was the dominant non-Saccharomyces species under all conditions, and M. pulcherrima was the minority species, being detected by PMA-qPCR throughout fermentation but at lower and decreasing concentrations. The use of a multistarter culture consisting of several non-Saccharomyces species previously selected for their ability to reduce ethanol and a S. cerevisiae strain resulted in reduced-alcohol wines, even if no aeration was applied. The sequential fermentations obtained ethanol reductions from 0.48-0.77% (v/v), which were accompanied by increases in the lactic acid and acetic acid contents. Among all fermentations, the highest reduction was obtained in the sequential inoculation with a higher inoculum size (10 7 cells/mL), when S. cerevisiae was added at 72 h. Nevertheless, the ethanol reduction obtained when using a multistarter of non-Saccharomyces species was lower than that when using each non-Saccharomyces species separately [39], which indicates no additive effect among them but a lower efficiency to deliver this outcome due to microbial interactions. Although not as efficient in reducing ethanol, a multistarter inoculation strategy could have additional benefits, such as improving the aroma profile and overall complexity of wines [23,32]. Nevertheless, the use of several species can also be challenged by winemaking conditions and the initial yeast population; therefore, special care has to be taken in the winery for this kind of procedure to be applied [32,77].

Data Availability Statement:
The data presented in this study are available on request from the corresponding author.

Acknowledgments:
The authors would like to thank Braulio Esteve-Zarzoso for support on the qPCR design.

Conflicts of Interest:
The authors declare no conflict of interest.