In this study, 59 yeast strains belonging to genera Saccharomyces [(S. eubayanus, S. kudriavzevii and S. uvarum, previously describes as cryotolerant), S. cerevisiae as a good fermentative species] and Kluyveromyces, Torulaspora, and Metschnikowia (Sipiczki et al., 2018) as controls of other species used in fermentation, were used. All the strains employed in the study are detailed in Supplementary Table S1. The industrial strains were kindly provided by Lallemand Inc. (France). These strains were codified as ADY and were named from ADY1 to ADY59. Inocula were prepared by introducing one single colony from the pure cultures of each strain into 5 mL of the same medium to be used in the experiments (YPD, LM, SM or NM). After a 24-h incubation of the precultures at 28°C, the volume required to obtain a concentration of about 10⁶ cells mL^-1 was inoculated in different media, as described below. The correct inoculation size was always confirmed by the surface spread on YPD agar plates.

The growth media selected for the experiments were YPD (glucose 20 g L^-1, peptone 20 g L^-1, yeast extract 10 g L^-1), a mineral media derived from that described by Verduyn et al. (1990), hereafter referred to as the lab medium (LM), and synthetic grape must (SM). The latter was derived from synthetic grape must (pH 3.3), as described by Riou et al. (1997), but with 200 g L^-1 of reducing sugars (100 g L^-1glucose and 100 g L^-1 fructose). The following were utilized: organic acids, malic acid 5 g L^-1, citric acid 0.5 g L^-1 and tartaric acid 3 g L^-1; mineral salts KH₂PO₄ 750 mg L^-1, K₂SO₄ 500 mg L^-1, MgSO₄ 250 mg L^-1, CaCl₂ 155 mg L^-1, NaCl 200 mg L^-1, MnSO₄ 4 mg L^-1, ZnSO₄ 4 mg L^-1, CuSO₄ 1 mg L^-1, KI 1 mg L^-1, CoCl₂ 0.4 mg L^-1, H₃BO₃ 1 mg L^-1 and (NH₄)₆Mo₇O₂₄ 1 mg L^-1; vitamins myoinositol 20 mg L^-1, calcium pantothenate 1.5 mg L^-1, nicotinic acid 2 mg L^-1, chlorohydrate thiamine 0.25 mg/L, chlorohydrate pyridoxine 0.25 mg L^-1 and biotin 0.003 mg L^-1. The assimilable nitrogen source used was 300 mg L^-1 (120 mg L^-1 as ammonium and 180 mg L^-1 in the amino acid form). The sporulation medium was KAc (potassium acetate 1%, agar 2%).

Growth was monitored by determining optical density at 600 nm in a SPECTROstar Omega instrument (BMG Labtech, Offenburg, Germany). Measurements were taken every 30 min for 4 days after 20-s of pre-shaking for the 25–40°C experiments, and every 90 min for 7 days for the low-temperature assays. Microplate wells were filled with the required volume of inoculum and 0.25 mL of the YPD or SM medium to always ensure an initial OD of approximately 0.1 (inoculum level of about 10⁶ cells mL^-1). For each experimental series, the non-inoculated wells were also included in the microplate to determine, and to therefore subtract, the noise signal (García-Ríos et al., 2014). All the experiments were carried out in triplicate. Growth parameters were calculated from each treatment by directly fitting OD measurements vs. time to the reparametrized Gompertz equation proposed by Zwietering et al. (1990):

where y = ln(OD_t/OD₀), OD₀ is the initial OD and OD_t is the OD at time t; D = ln(OD_t/OD₀) is the asymptotic maximum, μ_max is the maximum specific growth rate (h^-1) and λ is the lag phase period (h) (Aguilera et al., 2007). Overall yeast growth was estimated as the area under the OD vs. time curve (70 h and 168 h at 28°C and 15°C, respectively). This parameter was calculated by integration using the OriginPro 8.0 software (OriginLab Corp., Northampton, MA, United States).

The best 10 strains, selected for their μ_max at 15°C, were inoculated together in a flask to reach an OD_{600 nm} of approximately 0.2 in total (0.02 OD_600nm per strain) in order to mimic the population size used in the wineries and to follow a complete growth curve. Cultures were allowed to grow through a normal growth curve, with a transfer of a small volume (the volume required to once again inoculate at an OD_600nm of 0.2) of the expanded culture in 60 mL of fresh medium every 3 days. Culture growth was monitored by measuring absorbance at 600 nm every 24 h. After each transfer, cultures were plated on solid YPD and 50 colonies of each point were randomly selected and genotyped by a mitochondrial DNA restriction analysis (Querol et al., 1992).

The strains with the highest cell percentage in the competition experiments were selected to construct hybrids with other strains. All the selected strains were grown on 15 mL of YPD for 5 days at 28°C. Aliquots of each culture were seeded onto α-aminoadipic (α-AA) and 5-fluoroanthranilic acid (5-FAA) agar plates to select the lys- and trp- natural mutant colonies, respectively (Zarett and Sherman, 1985; Toyn et al., 2000). In order to confirm the presence of auxotrophy, the colonies that were able to grow were once again plated on new plates of MM [Yeast Nitrogen Base (YNB, Difco)], supplemented with 20 g L^-1 of glucose as the carbon source and with 5 g L^-1 of ammonium sulfate as the nitrogen source), α-AA and 5-FAA for 48–72 h at 28°C following the methodology proposed by Zarett and Sherman (1985), and partially modified by Pérez-Través et al. (2012).

Rare-mating assays were carried out according to the procedures proposed by Spencer and Spencer (1996), with some modifications (Pérez-Través et al., 2012). The strains carrying the auxotrophic markers were grown separately in 25 mL GPY broth for 48 h at 28°C. Cells were recovered by centrifugation (4,000 × g for 5 min at room temperature), and the pairs of yeast cultures to be hybridized were placed together in the same tube. Aliquots of these mixed strains were inoculated in 2 mL of fresh YPD medium. After 5–10 days of static incubation in the slanted position at 28°C, cells were recovered by centrifugation (4,000 ×g for 5 min at room temperature), washed in sterile water, re-suspended in 1 mL of PBS and incubated for 2 h. A heavy suspension of the mixed culture was spread on the MM plates and incubated at 28°C. Prototrophic colonies usually appeared after 3–5 days. These colonies were isolated and purified by restreaking on the same medium (MM). The hybrid nature was confirmed by the PCR amplification of the MAG2 and GSY1 protein-encoding nuclear genes, and the subsequent RFLP analysis with restriction enzymes MspI and TaqI (FastDigest, Thermo Scientific) (González et al., 2007).

The stabilization process was done by yeast sporulation. This process was induced by incubation on acetate medium for 5–7 days at 28°C. Following the preliminary digestion of the asci walls with 2 mg mL^-1 glucuronidase (Sigma), viability was calculated as the percentage of spores (from a total of 40 analyzed spores per hybrid strain) able to form a colony on YPD agar after 48–72 h at 28°C. The hybrid nature was confirmed by the PCR amplification of the MAG2 and GSY1 protein-encoding nuclear genes, and the subsequent RFLP analysis with restriction enzymes MspI and TaqI (FastDigest, Thermo Scientific) (González et al., 2007). Each selected spore was individually growed into 5 mL of YPD and incubated at 25°C. After reaching the stationary phase, an aliquot was used to inoculate a new tube. After five successive complete growths, an aliquot was plated on YPD plates and incubated at 25°C. Ten yeast colonies were randomly picked and characterized by inter-δ sequences (Legras and Karst, 2003) and RAPD-R3 (Pérez-Través et al., 2012) analyses. Simultaneously, the same colonies were inoculated in YPD and 10 colonies from each new culture were analyzed by the same methods. We considered a genetically stable spore when the colonies recovered after individual growths maintained the same molecular pattern than the previously inoculated (original) culture (Pérez-Través et al., 2012, 2014).

The DNA content of each parental, hybrid (H1) and segregant strains (S1-S3) was assessed by flow cytometry in a Beckman Coulter FC 500 (Beckman Coulter Inc., Brea, CA, United States) by the SYTOX Green dye method described in Haase and Reed (2002). The DNA content values were scored based on the fluorescence intensity compared with the S. cerevisiae diploid (FY1679) reference strain. The DNA content value reported for each strain was the result of two independent measures (Table 1).

Fermentations were performed at 28°C and 15°C with continuous orbital shaking at 100 rpm. Fermentations were done in laboratory-scale fermenters using 100 mL bottles filled with 80 mL of SM or natural grape-must (NM). Dimethyl dicarbonate (DMDC) at 1 mL L^-1 was added for sterilization purposes in NM. The NM used was Merseguera must, whose content was 173.45 g L^-1 fermentable sugar, and nitrogen levels were 182 mg L^-1. Flasks were closed with stoppers and airlocks to release CO₂. Fermentation was monitored by mass loss until a constant mass was reached, considered to be the end of fermentation. Experiments were carried out in duplicate. Yeast cells were removed by centrifugation and supernatants were stored at -20°C until further analyses. Potential contaminations were monitored with a negative control (NM without cells). The monitored mass loss was corrected to the percent of sugar consumed, as in Pérez-Través et al. (2015):

where C is the percent of sugar consumed at each time point, m is the mass loss value at that sampling time (g), S is the initial sugar concentration in the must (g L^-1), R is the final sugar concentration in the fermented must (residual sugar, g L^-1) and mf is the total mass loss value at the end of fermentation (g). The residual sugar values were obtained by the HPLC analysis of the fermentation samples. Curve fitting was carried out using the reparametrized Gompertz equation proposed by Zwietering et al. (1990).

Extracellular glucose, fructose, glycerol and ethanol were analyzed at the end of the SM and NM fermentations. Analytical HPLC was carried out in a Surveyor Plus Chromatograph (Thermo Fisher Scientific, Waltham, MA, United States) equipped with a refraction index detector, an autosampler and a UV–Visible detector. Prior to injection, samples were centrifuged at 13,000 rpm for 5 min, and then diluted 10-fold and filtered through 0.22 μm pore size nylon filters (Micron Analitica, Spain). A total volume of 25 μL was injected into a HyperREZ XP carbohydrate H + 8 mm column (Thermo Fisher Scientific) assembled to its correspondent guard. The mobile phase was 1.5 mM H₂SO₄ with a flux of 0.6 mL min^-1 and a column temperature of 50°C. The concentration of each compound was calculated using external standards. Each sample was analyzed in duplicate.

HPLC data were analyzed by the Statistica 7.0 software package (StatSoft, Tulsa, OK, United States) by a one-way ANOVA and a Tukey test for the means comparison. The cytometry results were tested by a one-way ANOVA and a Tukey HSD test (α = 0.05, n = 2). Phenotypic data were fitted to the reparametrized Gompertz model by non-linear least- squares fitting using the Gauss–Newton algorithm as implemented in the nls function of the RStudio statistical software, v.1.1.456. Dendrograms were created using the hclust package in the RStudio statistical software, v.1.1.456.

A phenotypic analysis was performed at different temperatures (12–40°C) to evaluate the thermotolerance differences of a collection of yeast strains belonging to diverse environmental niches and species. For this purpose, two different media, synthetic must (SM) and a complete laboratory medium (LM), were used. Figure 1 shows the maximum specific growth rate (μ_max) of the complete set of yeasts for the whole range of assayed temperatures. For these temperatures, the average of the optimum growth temperature of all these strains could be fixed at around 33°C for both media. In addition, the higher the temperature was, the wider variance became.

Figure 2 represents the phenotypic clustering of the complete set of strains assayed and grouped in growth behavior terms at all the tested temperatures. In both media, we observed two major groups (1 and 2) with some differences in their strain compositions. In SM (A), group 1 was integrated by all the studied S. cerevisiae strains and one strain belonging to K. marxianus. Group 2 was a mixture formed with the other study strains. Regarding LM (B), a dendrogram divided the strains once again into two groups (1 and 2) but, in this case, the separation between cerevisiae, non-cerevisiae and non-Saccharomyces was not as obvious as in SM. Group 1 was integrated mainly by all the S. cerevisiae strains, but some strains of genera Torulaspora and Kluyveromyces clustered together with them in this group. Thus the most homogenous group was subcluster 1.1, integrated by all the wine strains of S. cerevisiae. This phenotypic grouping is clearly determined by a combined effect of the growth medium and temperature. However, to the best of our understanding, the medium effect is stronger in SM, which separated S. cerevisiae, the most competitive species in a high-sugar medium such as SM, of the remainder strains, belonging to less competent species in this medium. In fact, the top 10 strains with higher μ_max at 15°C in SM belonged to S. cerevisiae (except Kluyveromyces spp. ADY42; Supplementary Table S2). Conversely, LM seems to be a less stressful medium, in which growth is mainly determined by the fitness at different temperatures. In LM, Cluster 1 was mainly integrated by the strains able to grow at 40°C (Supplementary Table S2), that is, the most tolerant strains at high temperature.

When the μ_max was calculated by species (Figure 3), practically all the strains showed a similar growth rate at 12–15°C and with very narrow differences up to 25°C. However, the temperature increases above 25°C provoked the greatest differences between S. cerevisiae and the remainder species, mainly in SM. This result is very interesting because demonstrates that the high temperature could be a more determining factor in the well-known capacity to outcompete to other species than the ethanol tolerance, as Salvadó et al. (2011a) already concluded. Recently, some species of non-Saccharomyces have drawn the attention of winemakers, as they positively modify the wine chemical composition, and consequently, improve their flavor and bouquet. Nowadays, different strains of K. marxianus, M. aff. pulcherrima and T. delbrueckii are available in the market to be co-inoculated or sequentially inoculated with S. cerevisiae strains. Our data evidenced that low temperature fermentations could be a good approach to increase their survival and contribution to the final wines.

In order to validate the competitiveness of the best-adapted strains at low temperature, we selected the top 10 strains showing higher μ_max values of the phenotyping done at 15°C in LM (Supplementary Table S2). As mentioned above, we selected LM to avoid any interference from other stresses, such as low pH, high sugar content and high ethanol, which happened with the SM medium. The best 10 strains belonged to species S. uvarum, S. kudriavzevii, S. cerevisiae, and T. delbrueckii. This latter species was represented by the strain T. delbrueckii ADY35 that showed the best μ_max of all the tested strains at 15°C. However, as our aim was to select the best parental for crossing with S. cerevisiae strains, we only selected the best 10 strains belonging to Saccharomyces genus for the competition experiment (Figure 4). The strain dynamic in the subsequent batch-cultures was analyzed by the restriction of the mitochondrial DNA of the isolated colonies in the different competition experiment stages. The result of this strain imposition was the presence of two S. uvarum strains (ADY57 and ADY59; Supplementary Table S1) throughout the process, which co-existed in the culture, with percentages of 45 ± 2.5 and 55 ± 1.4%, respectively. ADY57 was isolated from Tokaji wine in Hungary, while ADY59 (Rodríguez et al., 2014; Origone et al., 2018) was isolated from apple Chicha fermentation in Argentina. It is noteworthy that the two Saccharomyces strains that outcompeted the others were those with the highest μ_max values at 15°C (Figure 4). The competition experiment was also run at 28°C as the control condition and the yeast that reached 100% of cells in culture was the commercial S. cerevisiae strain ADY16.

We selected the spontaneous auxotrophic mutants of S. cerevisiae ADY18 (trp^-) and S. uvarum ADY59 (lys^-) as parental strains to perform hybridization to obtain a new hybrid strain with improved fermentation capacity at low temperature. ADY18 is a S. cerevisiae strain with very good enological features, but with the potential for improving its cryotolerance. S. uvarum ADY59 was the best strain in terms of μ_max and competitiveness at 15°C. After the rare-mating assay, the putative hybrids (prototrophic colonies) were checked by PCR and the positive ones were sporulated for a quick stabilization of its genome. After sporulation, the hybrid nature of the viable spores was again confirmed by the PCR amplification of the MAG2 and GSY1 protein-encoding nuclear genes, and the subsequent RFLP analysis with restriction enzymes MspI and TaqI. In this process, we obtained three viable spores (S1–S3) out of 40 (7.5% of viability). The stable status of the spores was tested by inter-δ sequences and RAPD-R3 after successive rounds of batch-cultures. DNA content analysis showed that S1–S3 were 3n (likely alloaneuploid) derived from a tetraploid hybrid (H1). This ploidy of the hybrid confirmed us that it was formed by a rare-mating event and not by a spore-spore cross. Moreover, we also checked that the parental strains were not able to sporulate under the rare-mating assay.

As a first selection step, all the stable spores (S1–S3), along with the two parental strains, were evaluated for growth parameters (Figure 5) and fermentative features (Table 2) in SM. Figure 5 shows the μ_max (Figure 5A) and the area under the curve (Figure 5B) of the three segregant spores compared with that of the parental strains at 15°C. The three segregants showed more than double the growth rate comparing with the S. cerevisiae (Sc) parental, and a slightly better value compared with the S. uvarum (Su) parental. The area under the curve (AUC) parameter allowed us to estimate overall behavior in all the growth phases. In relation to this parameter, hybrids obtained similar values compared with that of the Su parental, which almost doubled the Sc parental AUC value. As for the time needed to consume 50% and 100% of the sugars present in the must (Table 2), it should be noted that the three hybrid segregants showed lower T50 and T100 than both parentals at 15°C. The biggest differences between the Sc parental and segregants were observed in T50 because these spores have a much shorter lag phase than the industrial Sc strain (data not shown). At 28°C, segregants presented a similar fermentation kinetics compared with the Sc parental, while the Su parental presented a significantly delayed fermentation process as it was unable to consume the total amount of sugars and produced a stuck fermentation.

To confirm the data obtained in SM, fermentations were also carried out in the natural grape must of the Merseguera white variety. As in SM, the kinetic growth curves of the NM fermentations were inferred by following mass loss. Growth curves were used to extract the kinetic parameters of each strain (Table 2). The Merseguera fermentations performed by segregants at 15°C were faster than both parentals. These differences were especially greater with the Su parental, which required around 47 h more to finish the process. According to its cryotolerance, this strain (Su) started fermentation very quickly, with the shortest lag phase and the smallest T50. However, it was more sensitive to the increasing ethanol concentration and showed a delay in the last fermentation stages. The segregant strains (S1–S3) also had a shorter lag phase than the Sc parental, but sustained good fermentation activity by the end of fermentation, perhaps as a consequence of improved ethanol resistance in comparison with the Su parental. At 28°C, S1 and S3 also showed a faster fermentative kinetics compared with both parentals. The differences among the three segregants were noteworthy, and we highlight the importance of each parental’s DNA distribution. The concentrations of glucose, fructose, glycerol and ethanol at the end of the SM and NM fermentations are shown in Supplementary Table S3.