Wine Yeasts Selection: Laboratory Characterization and Protocol Review

Wine reflects the specificity of a terroir, including the native microbiota. In contrast to the use of Saccharomyces cerevisiae commercial starters, a way to maintain wines’ microbial terroir identities, guaranteeing at the same time the predictability and reproducibility of the wines, is the selection of autochthonous Saccharomyces and non-Saccharomyces strains towards optimal enological characteristics for the chosen area of isolation. This field has been explored but there is a lack of a compendium covering the main methods to use. Autochthonous wine yeasts from different areas of Slovakia were identified and tested, in the form of colonies grown either on nutrient agar plates or in grape must micro-fermentations, for technological and qualitative enological characteristics. Based on the combined results, Saccharomyces cerevisiae PDA W 10, Lachancea thermotolerans 5-1-1 and Metschnikowia pulcherrima 125/14 were selected as potential wine starters. This paper, as a mixture of experimental and review contributions, provides a compendium of methods used to select autochthonous wine yeasts. Thanks to the presence of images, this compendium could guide other researchers in screening their own yeast strains for wine production.


Introduction
Wine reflects the specificity of a terroir, including the microbial terroir [1]. This is particularly true in spontaneous fermentation by native wine yeasts that nevertheless expose winemakers to well-known complications. Most winemakers prefer to make use of commercial Saccharomyces cerevisiae starters, which guarantee predictability and reproducibility of the wines. On the other hand, the extensive use of worldwide-distributed commercial starters leads to organoleptic flattening and uniformization of the wines. Moreover, the positive contribution of non-Saccharomyces strains to must fermentation is well established. Non-Saccharpmyces species are known to modulate the wine aromatic profile in particular via esterase and β-glucosidase activities, but also to increase the glycerol content, to lower the alcohol content, and to exert proteolytic and pectinolytic activities that lead to enrichment of the aroma profile [2][3][4][5][6][7][8]. Besides S. cerevisiae commercial starters, non-Saccharomyces commercial starters have become available in recent years [9].
An alternative for preserving the role of the microbial terroir is to isolate and select autochthonous Saccharomyces and non-Saccharomyces strains towards optimal enological characteristics for use as co-or sequential inocula for wine production in the area from which they were isolated.
Enological characteristics to consider in the selection process are divided into technological and qualitative traits [10]. Technological traits (fermentation vigor, ethanol tolerance, The strains were grown on YPD agar at 30 • C for 2-3 days. Subsequently, a loopful from one colony of each strain was dispersed in distilled water, used to load the 96-position ZnSe plate, and dried at 37 • C for 45 min and spectra were measured with a Tensor 27 FTIR spectrometer (Bruker Optics, Ettlingen, Germany) using 32 scans per sample [40].

Yeast Screening
Depending on the test to perform, strains were grown overnight at 25 • C either on YPD agar or in YPD broth and then the cultures were used to inoculate either YPD broth or specific media. For the latter, each strain was harvested by centrifugation (5000 rpm for 10 min), washed once in NaCl 0.9% (w/v) solution, re-suspended to Optical Density (OD) 600 of 1.0 in the same solution [41]. Subsequently, each strain was spotted (5 µL) in duplicate onto specific media.

Macroscopic and Microscopic Observation, Type of Growth, CO 2 Production
Strains were grown in tubes containing YPD broth for 48 h at 25 • C to assess the modality of growth, and in YPD broth tubes with Durham tubes for 7 d at 25 • C to assess fermentation through CO 2 trapped in the Durham tubes. In addition, the strains were streaked onto YPD agar in order to record the colony morphology. A small quantity of the biomass was observed under a microscope (Olympus BX53, Tokyo, Japan) to record the cell morphology.

Spore Formation
The yeast biomass, taken with a sterile 1 µL loop, was streaked onto sodium acetate (1 g/L) agar (20 g/L) plates to check spore formation after 10 days of incubation at 25 • C [42].

Growth at Various Temperatures
Each strain pre-grown overnight at 25 • C was inoculated at 1% in YPD broth and statically incubated at both 18 • C and 37 • C for 24 h to test its ability to grow at low and high temperatures [24].

Low pH, Ethanol, and Osmotic Tolerance
Each strain culture was spotted onto YPD agar adjusted either to pH 3.0 or supplemented with 300 g/L of glucose or ethanol content (EtOH) of 5%, 10%, 12% or 15%. The plates to test the ethanol tolerance were freshly poured and sealed with Parafilm to prevent evaporation. All inoculated plates were incubated at 30 • C and colony development was checked daily [24].

SO 2 Resistance
Each strain culture was spotted onto YPD agar adjusted to pH 3.0 and supplemented with various concentrations of potassium metabisulphite [29]. A potassium metabisulphite stock solution was filter-sterilized (pore-size, 0.45 µm) and then added to the medium in concentrations of 80 (only for the strains selected for microvifications), 100, 200, 300, and 400 mg/L in order to correspond to half of the concentrations of SO 2 . The plates were incubated at 30 • C and the colony development was checked daily.

Catalase Activity
The yeast biomass, taken with a sterile 1 µL loop, was added to a drop of 3% (v/v) H 2 O 2 [43]. The development of bubbles indicated positive activity.

Acetic Acid Production
A loopful (1 µL) of biomass of each strain was streaked onto Chalk agar (yeast extract 3 g/L, glucose 10 g/L, calcium carbonate 3 g/L, agar 15 g/L) plates and incubated for 7 d at 25 • C [44]. The presence and extent of a clear halo around the yeast biomass indicated the rate of acetic acid production.

H 2 S Production
A loopful (1 µL) of biomass of each strain was streaked onto BiGGY agar plates and incubated for 48 h at 25 • C [45]. The color intensity of the biomass indicated the rate of H 2 S production [20].

β-Glucosidase Activity
Each strain culture was spotted onto Petri plates containing arbutin (5 g/L), yeast extract (10 g/L), 40 drops/100 mL of a 1% solution of ferric ammonium citrate solution, and agar (20 g/L) according to Caridi et al. [46]. After incubation at 25 • C for 7 days, this activity was indicated by the medium changing color, from pale to dark brown.

Pectinase Activity
Each strain culture was spotted onto Petri plates with YNB (6.7 g/L), apple pectin (12.5 g/L), and agar (10 g/L), adjusted to pH 4 with 1 N HCl according to Sidari et al. [47]. After 10 days of incubation at 25 • C, activity was determined by measuring the diameter of the colonies and checking for the presence of a clear halo after flooding the plates with Lugol's solution and washing with water [27].

Esterase Activity
Each strain culture was spotted onto Petri plates with peptone (10 g/L), NaCl (5 g/L), CaCl 2 ·2H 2 O (0.1 g/L), Tween 80 (10 g/L), and agar (20 g/L) at pH 6.8 [48]. After 6 days of incubation at 25 • C, the ability to hydrolyze esters was estimated by the presence of a visible opaque precipitate around the colony.

Protease Activity
Each strain culture was spotted onto Petri plates with a medium prepared by mixing the two following solutions: malt extract 3 g/L, yeast extract 3 g/L, peptone 5 g/L, glucose 10 g/L, NaCl 5 g/L, agar 20 g/L (separately sterilized), adjusted to pH 3.5 with 0.1 M HCl; and a skim milk solution (10% w/v) prepared and treated at 100 • C for 10 min. After incubation for 3 days at 25 • C, the presence of a clear halo around the yeast spot indicated protease activity [29].

Micro-Fermentation Trials in Grape Must
Eleven out of 29 strains, chosen considering the results of the above-reported screening tests, were tested for fermentation vigor after 2 d and 7 d in Trauben saft-100% Direktsaft red grape must (dmBio; Drogerie Markt, Wals-Siezenheim, Austria), both with and without SO 2 supplementation. The sugar content of the must was 16 • brix and the pH was 3.20. Aliquots of 100 mL of the must were distributed into flasks, 10 mL of liquid paraffin was added to avoid surface contact with oxygen, and the resulting mixtures were pasteurized at 100 • C for 20 min. Subsequently, half of the flasks were supplemented with 80 mg/L of potassium metabisulphite. All flasks were inoculated in duplicate with 5 mL of 48 h pre-cultures grown in the same red grape must incubated statically at 25 • C. Fermentation progress was monitored by recording weight loss due to the release of CO 2 . After 2 d and 7 d, fermentation vigor was expressed as g of CO 2 /100 mL of must [20].
At the end of the fermentation, the wines produced with and without SO 2 were analyzed for pH, total titratable acidity (TTA), volatile acidity, ethanol, glucose and fructose, total polyphenols, total flavonoids, and volatile organic compounds (VOCs).
The parameters of ethanol, TTA and volatile acidity were determined according to the official methods of International Organisation of Vine and Wine (OIV) [49]. The results were expressed as the mean of four determinations ± standard deviation.
Determination of glycerol and glucose and fructose sugars was accomplished via high performance liquid chromatography (HPLC) using a PU-4003 chromatograph (Pye Unicam, Cambridge, UK) in accordance with the accredited method published by Suhaj and Belajová [50]. Chromatographic separation was performed on the Kromasil 100-5NH2amino column, 250 × 4.6 mm i.d. (EKA Chemicals AB, Sweden), using an RID-10A refractive index detector (Shimadzu, Tokyo, Japan). The RID optical unit was permanently warmed to 40 • C. All wine samples were injected into 20 µL volumes and eluted isocratically with mobile phase acetonitrile:water, 80:20 (v/v). The flow rate of the mobile phase was 1.5 mL/min. The peaks were identified by retention times and quantified via external calibration using the software QC Expert version 2.5 (TriloByte Statistical Software, Pardubice, Czech Republic). Wine samples were diluted twofold with deionized water and filtered through a 0.45 µm syringe filter with a cellulose membrane (Agilent, Waldbronn, Germany) and subsequently injected for HPLC. During the calibration measurements the correlation coefficients were higher than 0.98 for all analyzed compounds. The refractive index detector responses were linear in the range of 0.4-20 g/L for glycerol and 0.5-50 g/L for glucose and fructose.
Total polyphenols (TPC) and total flavonoids (TFC) were determined using a Shimadzu 3600 UV-VIS-NIR spectrophotometer (Shimadzu, Tokyo, Japan) with an accessory. All experiments were performed in duplicate. A 12% (v/v) ethanol was used as a reference.
TPC was determined by applying the Folin-Ciocalteu modified method [51]. Briefly, 100 µL of wine sample was appropriately diluted with 12% (v/v) ethanol, 7.9 mL of distilled water, and 500 µL of Folin-Ciocalteu reagent and mixed in a 20 mL vial. After 10 min, 1.5 mL of 20% sodium carbonate was added, and the contents mixed. Samples were incubated at room temperature in darkness for 60 min, and absorbance was measured at 765 nm. Standard solutions of gallic acid were used to construct the calibration curve (0-1500 mg/L). The results were expressed as gallic acid equivalent (GAE, mg/L).
TFC was evaluated according to the modified method with aluminum chloride [51]. Briefly, 500 µL of wine sample was added to a 10 mL vial containing 1.5 mL of 96% ethanol and 2.8 mL of distilled water. After this, 100 µL of 10% aluminum chloride and 100 µL of 1 M potassium acetate were added, and the contents mixed. After 40 min, the absorbance of the final solution was measured at 415 nm. Standard solutions of quercetin were used to construct the calibration curve. The results were expressed as quercetin equivalent (QE, mg/L).
To analyze VOCs, solid phase micro-extraction (SPME) was carried out using a polydimethylsiloxan-divinylbenzene fiber, coating thickness 65 µm (Supelco, Bellefonte, PA, USA), immersed in 10 mL of wine sample and mixed at 6 Hz on a magnetic stirrer during 30 min at 20 • C. The extracted compounds were analyzed via gas chromatographymass spectrometry (GC-MS) using a 6890N gas chromatograph (Agilent Technologies, Santa Clara, CA, USA) coupled to a 5973 mass spectrometric detector (Agilent Technologies). The SPME fiber was placed in the inlet of the chromatograph for 2 min at 250 • C so as to desorb the extracted compounds. The gas chromatographic separation took place in a DB-WAXetr high polarity polyethylene glycol column (length 30 m, inner diameter 0.25 mm, stationary phase thickness 0.5 µm; Agilent Technologies) using a temperature program of 35 • C for 1 min, 5 • C for 1 min and 250 • C for 1 min. The split ratio was 10:1. The average velocity of the He carrier gas was 34 cm·s -1 at constant flow. An ionization voltage of 70 eV was used. Compounds were identified by comparison of mass spectra with the NIST 14 MS library (National Institute Standards and Technology, Gaithersburg, MD, USA). Analysis was carried out solely at orientation level with relative quantification data expressed as peak area percentage.

Statistical Analysis
Data were analyzed using a one-way ANOVA and Tukey's test at 5% probability level, using the online tool at https://www.statskingdom.com/doc_anova.html (accesed on 9 March 2021). Table 2 reports the grouping of strains via 5.8-ITS rRNA analysis and RFLP and their molecular identification by sequencing and comparison with the GenBank database.

Yeast Strains Grouping and Identification
The ITS amplicons had sizes ranging from 380 to 800 bp. Eighteen RFLP patterns were observed; different profiles were also assigned to strains belonging to the same species (H. uvarum, L. thermotolerans, M. pulcherrima, P. fermentans, T. delbrueckii, S. cerevisiae).
From among the twenty-nine yeast isolates, five were identified as M. pulcherrima, five as T. delbrueckii, five as S. cerevisiae, three as H. uvarum, three as P. fermentans, two as L. thermotolerans, one as Candida dubliniensis, one as Debaryomyces hansenii, one as Metschnikowia aff. chrysoperlae, one as Meyerozyma guilliermondii, one as Pichia kluyveri, and one as Zygosaccharomyces bailii.
The accession numbers of the yeast strains sequenced and deposited to GenBank are: MZ207954 C. dubliniensis CCY 29-178-1, MZ207959 D. hansenii 5-1-6, MZ207966 H.  Figure 1 reports the clustering of the strains obtained from the FTIR analysis. The mid-infrared range of 4000-500 wavelength/cm 2 (25,000-2500 nm) is used to excite atoms in molecular bonds, causing them to vibrate. A spectrum can be measured and calculated by light absorption. This specific absorption is then attributed to cell components (e.g., polysaccharides, fatty acids, proteins, mixed region, fingerprint region) [52] used for identification [53]. To enhance the resolution of complex bands and to minimize difficulties evolving from inevitable baseline shifts, the second derivations of the original spectra were calculated. This made it possible to obtain a list of the most similar spectra from the database [54], leading to identification at the species level.
Isolates were classified into sub-clusters by defining a spectral distance as a value for separation on the strain level. The spectral distance chosen was 0.1. The grouping reported in the figure closely corresponds to the results obtained from sequencing.    Table 3 reports the morphology, color, and texture of the colonies of the different strains grown on YPD agar. The strains belonging to the Metshnikowia genus exhibited biomass turning red during the prolonged incubation time. The microscopic morphologies of strains belonging to the species chosen as representative are reported in Figure 2.

Yeast Screening
All strains grew in YPD broth as dispersed cells, with the exception of the strain M. pulcherrima CCY 69-2-15, which exhibited growth with aggregated cells.

Yeast Screening
All strains grew in YPD broth as dispersed cells, with the exception of the strain M. pulcherrima CCY 69-2-15, which exhibited growth with aggregated cells.
Concerning growth at 37 • C, the strains T. The ability of the strains to grow under the stressed conditions possibly occurring during must fermentation was studied to consider their potential application in vinification. All strains were able to grow well in an osmotic stress condition (300 g/L of glucose) and at pH 3.0. Table 4 reports the biochemical activities of the twenty-nine strains tested.
Concerning ethanol tolerance after one day of incubation (Table 4a), all strains grew well in the medium supplemented with 5% of ethanol, while differences were observed with increasing ethanol concentration. In particular, with 10% of ethanol three strains grew very well-S. cerevisiae PDA M 1/1, L. thermotolerans 5-1-3, P. kluyveri PDA W 9. The strains that tolerated 12% of ethanol to various extents were those belonging to S. cerevisiae, L. thermotolerans, M. pulcherrima 125/14, P. kluyveri PDA W 9, Z. bailii 24-1-25, and two strains of T. delbrueckii. This trend was observed also in the presence of 15% of ethanol, with the exception of the strains P. kluyveri PDA W9 and Z. bailii 24-1-25, which did not grow.
After one day of incubation in media supplemented with increasing concentrations of SO 2 (Table 4a), S. cerevisiae strains showed good growth indicating full resistance (to 400 mg/L of metabisulphite), with the exception of the S. cerevisiae 60/16 which did not experience good growth with more than 200 mg/L of metabisulphite. This strain did show good growth in media supplemented with 300 and 400 mg/L of metabisulphite after two and three days of incubation, respectively. Non-Saccharomyces strains showed marked differences in a strain-dependent manner. The strains belonging to Lachancea, Debaryomyces, Hanseniaspora, and Meyerozyma genera exhibited no growth even at the lowest SO 2 concentration. Some of the strains needed longer incubations to confirm their incapability to grow in presence of SO 2 or to resist at different concentrations. As example, L. thermotolerans 5-1-1 grew after two days of incubation in the presence of 100 mg/L of metabisulphite, while at the highest SO 2 concentration only P. kluyveri PDA W 9 and the above mentioned S. cerevisiae 60/16 improved their growth over the incubation period.   All eleven strains chosen for the microvinification trials exhibited good growth in the presence of 80 mg/L of metabisulphite.
All strains were catalase-positive, with non-Saccharomyces strains exhibiting the highest activity, especially strains belonging to the species T. delbrueckii, D. hansenii, M. pulcherrima, H. uvarum, and P. fermentans (Table 4b).
Only 6.9% of the strains were positive for acetic acid production. These belonged to the species H. uvarum (Figure 4). and three days of incubation, respectively. Non-Saccharomyces strains showed marked differences in a strain-dependent manner. The strains belonging to Lachancea, Debaryomyces, Hanseniaspora, and Meyerozyma genera exhibited no growth even at the lowest SO2 concentration. Some of the strains needed longer incubations to confirm their incapability to grow in presence of SO2 or to resist at different concentrations. As example, L. thermotolerans 5-1-1 grew after two days of incubation in the presence of 100 mg/L of metabisulphite, while at the highest SO2 concentration only P. kluyveri PDA W 9 and the above mentioned S. cerevisiae 60/16 improved their growth over the incubation period.
All eleven strains chosen for the microvinification trials exhibited good growth in the presence of 80 mg/L of metabisulphite.
All strains were catalase-positive, with non-Saccharomyces strains exhibiting the highest activity, especially strains belonging to the species T. delbrueckii, D. hansenii, M. pulcherrima, H. uvarum, and P. fermentans (Table 4b).
Only 6.9% of the strains were positive for acetic acid production. These belonged to the species H. uvarum (Figure 4).  The strains exhibited a wide range of H 2 S production with the biomass color ranging from white to black and passing through intermediate tints ( Figure 5). The majority of the strains (41.38%) had hazel biomass followed by dark hazel (37.93%), black (13.79%), pale hazel (3.45%), and white (3.45%) (Table 4a). The strains exhibited a wide range of H2S production with the biomass color ranging from white to black and passing through intermediate tints ( Figure 5). The majority of the strains (41.38%) had hazel biomass followed by dark hazel (37.93%), black (13.79%), pale hazel (3.45%), and white (3.45%) (Table 4a). Thirty-one percent of the strains exhibited light to strong β-glucosidase activity ( Figure 6). The highest activity was recorded for strains belonging to the genus Metschnikowia and the strain P. kluyveri PDA W 9 (Table 4b). Thirty-one percent of the strains exhibited light to strong β-glucosidase activity ( Figure 6). The highest activity was recorded for strains belonging to the genus Metschnikowia and the strain P. kluyveri PDA W 9 (Table 4b). Figure 5. Different degrees of H2S production among yeast strains according to the color of the biomass; white: low production, hazel: medium production, dark: high production.

Enological Characterization by Micro-Fermentations
The eleven strains selected through the screening process were tested using microfermentations in order to evaluate their fermentation performance. The non-Saccharomyces strains showed lower fermentation vigor than the S. cerevisiae strains, which exhibited the highest fermentation vigor. Similar observations were reported for M. pulcherrima 125/14. The lowest values reported were for D. hansenii 5-1-6. Similar results were reported for fermentation vigor without and with SO2, confirming the results obtained by screening on plates and indicating the resistance of the strains to the used SO2 concentration (Table 5). The choice was made balancing the screening parameters' results for each strain while also taking into account the possibility of further improving their characteristics through different techniques, such as hybridization. In detail, the strains chosen exhibited no or very low acetic acid production, low to medium H 2 S production, good ethanol and SO 2 tolerance after one or two days, and a varied range of catalase, β-glucosidase, esterase, pectinase, and protease activities (Table 4).

Enological Characterization by Micro-Fermentations
The eleven strains selected through the screening process were tested using microfermentations in order to evaluate their fermentation performance. The non-Saccharomyces strains showed lower fermentation vigor than the S. cerevisiae strains, which exhibited the highest fermentation vigor. Similar observations were reported for M. pulcherrima 125/14. The lowest values reported were for D. hansenii 5-1-6. Similar results were reported for fermentation vigor without and with SO 2 , confirming the results obtained by screening on plates and indicating the resistance of the strains to the used SO 2 concentration (Table 5).  Tables 6-8 show the physicochemical parameters of the wines produced using the eleven selected yeast strains. pH, TTA, and volatile acidity values for the wines produced with each strain are reported in Table 6. All produced wines had a pH higher than the un-inoculated musts, and the TTA values were linearly correlated to pH. The pH range for the trials without SO 2 was 3.34-3.48, while for the trials with SO 2 it was 3.29-3.44. The volatile acidity ranged from 0.12 to 1.68 g/L of acetic acid in the absence of SO 2 . Table 6. pH, total titratable acidity, and volatile acidity of the wines produced by inoculating the red must without and with SO 2 for the eleven yeast strains.  Values represent mean ± SD from three measurements. Values in a column with identical superscript letters are not statistically different according to one-way ANOVA with Tukey's test at a statistical significance level of 0.05. n.d.: not detected. Analytical parameters for glucose determination were: limit of detection (LOD) = 0.08 g/L, limit of quantification (LOQ) = 0.10 g/L. Analytical parameters for fructose determination were: LOD = 0.04 g/L, LOQ = 0.06 g/L. Table 8. Glycerol, total polyphenol, and total flavonoids content of the wines produced by inoculating the red must without and with SO 2 for the eleven yeast strains.

Strains Glycerol (g/L) Total Polyphenols (GAE, mg/L) Total Flavonoids (QE, mg/L) w/o SO 2 with SO 2 w/o SO 2 with SO 2 w/o SO 2 with SO 2
Concerning ethanol production (Table 7), S. cerevisiae strains consumed glucose and fructose as expected, with produced ethanol percentages as high as 8.39% and 8.23% by S. cerevisiae 15-1-552 without SO 2 and S. cerevisiae PDA W10 with SO 2 , respectively. M. pulcherrima 11-1-7 fermented 33 and 30 g/L of glucose in micro-fermentation without and with SO 2 , respectively, and 23 and 19 g/L of fructose in micro-fermentation without and with SO 2 , respectively, producing the lowest percentages of ethanol (4.15% and 4.10% without and with SO 2 ). L. thermotolerans 5-1-1 fermented 80 g/L of glucose in trials without SO 2 and consumed almost all of the glucose in trials with SO 2 ; it fermented 70 and 81 g/L of fructose in the absence and presence of SO 2 , respectively, producing the highest values of ethanol (approximately 7.15-7.25%). Moreover, T. delbrueckii 3-16-1 produced wines with approximately 7% of ethanol. The lowest ethanol production was recorded for D. hansenii 5-1-6.
Comparing the total concentrations of polyphenols and flavonoids between the inoculated and un-inoculated musts (Table 8), in wines without SO 2 , increases or decreases in concentrations of polyphenol and flavonoid were reported in a strain-dependent manner. By contrast, in wines produced with SO 2 , a decrease in concentration of flavonoids was observed for all strains tested (Table 8).
During GC-MS analysis, various volatile aroma compounds were detected. Most of them are well known as contributors to wine aroma [58,59]. All fermented samples, with the exception of that fermented by D. hansenii 5-1-6, contained high amounts of 2phenylethanol, which is an established aroma compound with a sweet, floral, rosy character. The highest amounts of this compound were produced, among samples fermented with SO 2 (Table S1), by L. thermotolerans 5-1-1. Only samples fermented by S. cerevisiae strains, together with that fermented by M. pulcherrima 125/14, contained remarkable amounts of 4-vinylguaiacol, which is an aroma compound with a sweet-smoky character, typical for Traminer wines or for whisky. Various strains produced medium-chain fatty acids, a phenomenon more pronounced in samples with SO 2 (Table S1). Several samples contained considerable amounts of pyran and furan derivatives, which were probably sourced or metabolized from the UHT-treated substrate. Dodecanoic acid was detected only in various fermented samples treated with SO 2 , while only some fermented samples without SO 2 (Table S2) contained propylene glycol, 4-cyclopentene-1,3-dione, diethyleneglycol ethylether, 2-methylthiolane, hexanoic acid and nonanoic acid.

Discussion
This contribution aimed to give a guide as comprehensive as possible to methods for wine yeast selection while studying our own strains. We decided to test the strains for all the characteristics to obtain for each of them a complete profile in view of possible genetic improvement. The simple trials used allowed us to exclude those strains possessing the worst features (alone or in combination) for wine-making-high acetic acid and H 2 S production, low ethanol and SO 2 tolerance, foam production, zero or low enzymatic activity-selecting the best strains to test in must fermentations.
The strains here reported as M. pulcherrima are to be considered M. pulcherrima-like strains due to the difficulty in assigning an exact taxonomic position, as a result of a lack of distinctive morphological and physiological properties among species belonging to the M. pulcherrima clade and the lack of rDNA barcode gaps [60][61][62][63]. In contrast to Sipiczki [63], our strain of M. aff. chrysoperlae is a pigmented strain, as are all of our strains of M. pulcherrima. In our study, since it is known that for taxonomical proposes it is necessary to use more gene markers in order to well classify yeast strains, we chose to use only ITS fragment sequencing as an identification tool in combination with the RFLP and FTIR approaches. The sequencing of ITS regions is suitable as a rapid and preliminary identification tool for yeasts, which can then be deeply taxonomically analyzed exploiting other molecular markers as shown by previous studies [60][61][62][63].
FTIR spectroscopy facilitates the grouping of yeasts based on the chemical composition of their cells. It is a high-throughput method requiring no chemicals to be used, and therefore is cheap and convenient. Our results presented in Figure 1 demonstrate the overall success of this method to group yeast strains similarly to the sequencing-based approach, which is much more tedious and costly. Based on this, and based on our experience and several other studies [40,53,54], we can recommend the use of FTIR spectroscopy for preliminary grouping of strains and reducing the number of strains in order to pass to further evaluation, by elimination of those that are most probably duplicates or multiplicates.
The initial yeast concentration used in the screening might differ among yeast strains and species due to different morphology and size; therefore, we compared each strain with its own control condition, avoiding the comparison of different yeast species.
The presence of sulfur off-flavor in wine as a result of yeast metabolism is negatively correlated to wine quality, as it is an undesired wine off-flavor, and it also gives rise to health concerns. The screening of yeast strains that produce zero or low H 2 S is particularly necessary to take into account for the production of organic and sulfite-free wines. The degree of H 2 S production by non-Saccharomyces yeasts and the wide intra-species variability observed by different authors are consistent with the findings of this study [28,29,64]. By contrast, our findings for H. uvarum and M. pulcherrima conflict with the results of Polizzotto et al. [22] and Belda et al. [65] who reported absent or low sulfite-reductase activity.
The ability of the non-Saccharomyces strains under study to grow under stressed conditions was assayed to understand their potential application in vinification. The strains' ability to grow at low pH and in high concentrations of glucose makes them suitable for harsh environments; in addition, their growth at different temperatures makes them suitable for red and white vinification. Our results confirm that S. cerevisiae is the most ethanol-tolerant species compared to many non-Saccharomyces. Concerning the non-Saccharomyces strains, some of them exhibited higher ethanol tolerance (up to 12% or to a lesser extent up to 15%) compared to results from the literature [28,[62][63][64][66][67][68], confirming the results of Mukherjee et al. [69] for L. thermotolerans, T. delbrueckii and Z. bailii. The use of SO 2 in winemaking is mandatory to control spoilage and microorganisms and to protect wines from oxidation. Therefore, it is important for wine yeasts to be able to tolerate SO 2 at the dosage commonly used for commercial wine fermentation; on the other hand, the health aspect has to be taken into account. For this reason, although all strains were tested at increasing concentrations of SO 2 (100-400 mg/L), the microvinification trials were carried out using a low SO 2 concentration (80 mg/L). Yeasts belonging to Torulaspora, Metschnikowia, Zygosaccharomyces, and Lachancea genera were less sensitive to SO 2 than commonly considered [23] and this is consistent with other authors' results [28,29]. Concerning the strains' contribution to the pH of wine, our strain of L. thermotolerans confirms the existing strains' variability in producing lactic acid and consuming malic acid [66]. The selected L. thermotolerans strain did not significantly influence pH and total acidity compared to the S. cerevisiae controls (53, PDA W 10, PDA M1/1) ( Table 6). Previous studies report variability in lactic acid production from 0.2 g/L to approximately 10 g/L and reductions in pH from insignificant differences to 0.5 [70][71][72]. However, many other quality parameters can be improved by L. thermotolerans, so the choice of strain can be interesting. Results from the catalase test gave information on the ability of strains to cope with oxidative stress and to perform better during fermentation [73]. All of the strains tested in this study were catalase positive to various extents and in agreement with other authors for H. uvarum, Candida, Pichia, D. hansenii, M. pulcherrima [27,28].
The wine industry makes use of protease and pectinase to prevent wine haze and facilitate wine clarification, together with glycosidase to favor the expression of grape varietal aromas [74]. Wine yeasts can possess one or more natural enzymatic activities useful for vinification [32,61,[75][76][77]. Yeast enzymes of interest include esterases, glycosidases, proteases, and cellulases able to hydrolyze structural components [78,79] that determine, based on their presence and intensity, the sensorial complexity of wines [80].
Fermentation vigor is a good indicator of the strain's promptness and of the progress of the fermentation. It is easy to monitor as a weight measurement, which is directly proportional to sugar consumption and ethanol synthesis, determining the fermentation power. As expected, the S. cerevisiae strains had higher fermentation vigor than the non-Saccharomyces strains, in agreement with Caridi et al. 2002 [20], with the exception of the M. pulcherrima 125/14 strain. This strain, in fact, had higher fermentation vigor than the usually reported 4.5% ethanol (v/v) [89]. However, it is reported that some strains of M. pulcherrrima can produce 9-11.5% ethanol [90], with an ethanol tolerance of at least 6% with a few exceptions above 9% [28].
Acetic acid is one of the compounds that impact the sensory profile of wine, contributing to definitions of its quality. An acetic acid concentration of 0.7-1.1 g/L is considered unpleasant; the maximum acceptable limit for volatile acidity in most wines is 1.2 g/L of acetic acid [91,92]. Values in the range 0.2-0.7 g/L are usually considered optimal [91]. Although most non-Saccharomyces are considered high acetic acid producers [93,94] other evidence indicates T. delbrueckii, L. thermotolerans, M. pulcherrima as low producers [29,[95][96][97][98][99]. Our results were consistent with these reports, with all selected being within the optimal range with the exception of H. uvarum 26/17, the highest producer as reported for the species by Aponte and Blaiotta [99]. However, the behavior of this strain confirmed the utility of the visible halo as a screening test [16,77,100].
The quality of wine is also linked to the glycerol concentration, although this has recently been contested [101]. Noble and Bursick [102] indicated 5.2 g/L as the taste threshold with a maximum acceptable level of 25 g/L [103]. It is usually reported that glycerol production is higher in wines fermented with non-Saccharomyces compared to those produced with S. cerevisiae [101]. In addition, Zhu et al. [104] reported higher glycerol concentration for non-Saccharomyces than for S. cerevisiae, which was in agreement with the results of the majority of our tested strains. The role of this metabolite must be considered taking into account its relationship with ethanol and acetic acid production. In fact, higher glycerol production could result in a reduction in ethanol production [105] and a higher production of acetic acid [106].
The role of phenolics in wine is related to the sensorial and health aspects. The effect of SO 2 in vinification is well known [107] as is the contribution of S. cerevisiae to the polyphenolic profile of wine [108][109][110][111][112][113]. Recently, Morata et al. [114] reported different antocyanin adsorption by non-Saccharomyces with the goal of improving the color stability of wine. The different polyphenols concentrations in wines produced by our tested yeasts could be attributable to the strain (production of metabolites and cell wall adsorption) used in wines produced without SO 2 and to the concurrent role of strain and SO 2 in wines produced with the addition of SO 2 .
Analysis of volatile compounds using GC-MS aimed to determine the production of known and described aroma-active compounds by individual yeast strains. Although the aroma character of some of the compounds has been described as "pleasant" and their presence in wine is often appreciated, those described as "unpleasant" (or off-flavors) may be very important to achieve the required complexity, fullness and/or typicality of a wine aroma, when present at appropriate concentrations and in certain combinations. Lists of common aroma-active compounds in various types of wine are available in the literature, which facilitates their tracing in experimental samples. However, interpretation of the analytical data need not be straightforward and usually requires combination with sensorial evaluation of the wine bouquet [58,59].
Based on their performance, we propose S. cerevisiae PDA W 10, L. thermotolerans 5-1-1 and M. pulcherrima 125/14 as potential wine starters. Due to the characteristics of L. thermotolerans and M. pulcherrima, these yeasts are normally used in mixed or sequential fermentations together with S. cerevisiae [66,114] in order to complete the fermentation process, guaranteeing the quality of the wine. S. cerevisiae PDA W 10 could be used as pure inoculum. Moreover, after further studies on competitive abilities of the strains and examination of the killer trait of S. cerevisiae strains, L. thermotolerans 5-1-1, M. pulcherrima 125/14, and S. cerevisiae PDA W 10 could be used as co-or sequential inocula. It must be highlighted that the M. pulcherrima 125/14 strain has interesting properties, such as its fermentation vigor, that allows for the consideration of the use of this strain as pure inoculum as well.
The step-by-step process for screening and selecting wine yeasts is as follows: yeast isolation/revitalization of stored yeasts, yeasts identification, Petri plate screening for useful enological characteristics, evaluation of results and choice of strains, use of the chosen strains for micro-vinification, analyses of the produced wines, evaluation of results and identification of wine starter strains.

Conclusions
Even in the present era of modern molecular technologies, we believe it is important to maintain the know-how of classical methods for selecting wine strains useful for production of high-quality wine. The screening and testing procedures led to the selection of strains that could be the starting point for improvement by hybridization, mutagenesis, or genome engineering. Finally, we believe that this contribution reporting procedures and images could help other scientific groups in screening their own isolated yeasts for wine production.
Supplementary Materials: The following are available online at https://www.mdpi.com/article/10 .3390/microorganisms9112223/s1: Table S1. Volatile organic compounds of wines with SO 2 produced using the eleven yeast strains, Table S2. Volatile organic compounds of wines without SO 2 produced using the eleven yeast strains.