Biological material and experimental site

We used apples (Malus domestica Borkh) of the varieties ‘Golden Delicious’, ‘Red Delicious’, ‘Joya’, ‘Rayada’, ‘Royal Gala’, ‘Agua Nueva’, ‘Rosada’, ‘424’, ‘428’, ‘429a’, ‘429b’, ‘436’, ‘467’ and ‘Malus x Micromalus’ established in a phenological orchard in San José Itho, Amealco, Querétaro, Mexico, located at 20° 11' N and 100° 08' W, at 2,275 masl. The orchard is rain-fed, with auxiliary irrigation during the dry spring period. The average annual temperature is 15.2 °C, with average annual rainfall of 927.7 mm (^{Servicio Meteorológico Nacional [SMN], 2016}) and an average accumulation of chill hours of 500. The soil is deep (> 50 cm) and clayey, with pH 5.6, low salinity (0.03 dS·m^-1) and low organic matter content (0.53 %).

The reference yeast strains of Saccharomyces cerevisiae were K1-V1116 (LALVIN^®) and AH22 (sensitive to the killer genotype).

Yeast isolation

Apples were harvested according to their apparent maturity (visual appreciation of color) due to the scarcity of material; from these, the musts were obtained with a standard extractor (TU04, Turmix, Mexico). All musts were adjusted to 20 °Brix sucrose, sulphited (50 mg·L^-1 of SO₂), racked at 3 °C for 24 h and left to ferment spontaneously at 19 °C. Halfway through the fermentation (ρ = 1.03 g·mL^-1) and at its end (ρ = 0.99 g·mL^-1, or when the density remained constant), aliquots were taken, serial decimal dilutions were made, 100 μL were spread plated on nutrient yeast dextrose agar (NYDA) ‒supplemented with chloramphenicol (100 mg·L^-1) and rose bengal (60 mg·L^-1)‒ and incubated for two days at 28 °C.

From the plates corresponding to each apple variety, five colonies with contrasting morphologies were isolated at each stage of fermentation. The isolated strains were differentiated by three consecutive passes in lysine agar medium incubated at 25 °C for two days. Strains that did not grow in the medium were considered to be of the genus Saccharomyces (^{Medina et al., 2007}).

Saccharomyces yeast selection tests

To determine killer effect, the methodology proposed by ^{Lopes and Sangorrín (2010)} was used, with some modifications. First, a total of 6 Log of CFU·mL^-1 of the killer effect-sensitive strain (AH22) were poured plated in NYDA medium with methylene blue (0.2 %), NaCl (1 %) and phosphate citrate to adjust the pH to 4.6. Once the medium solidified, 5 μL of each strain were inoculated in drop form at a concentrationof 6 Log CFU·mL^-1 and incubated for 72 h at 25 °C. The strains were considered killer when a blue-framed inhibition halo was observed around them.

β-glucosidase activity was determined in esculin glycerol agar medium at pH 6, for which each strain was separately streak plated, and incubated at 25 °C for two days. β-glucosidase activity was evidenced by the hydrolysis of the substrate, in which a brown precipitate forms around the colony (^{Pérez et al., 2011}).

Flocculation capacity was determined as reported by ^{Suárez-Valles et al. (2008)}, with some modifications. The strains were inoculated in 5 mL of NYDB broth and incubated at 25 °C for 72 h. Subsequently, the cultures were centrifuged and resuspended in 5 mL of Helm’s buffer solution (3 mmol·L^-1 of CaCl₂ and 50 mmol·L^-1 of acetate-acetic buffer solution, pH 4.5). The flocculation index (FI) was determined by the ratio, in percent, of the optical density at 620 nm of the culture suspension and that obtained 120 min after adding Helm's solution. The strains were classified as: highly flocculant (FI < 30 %), moderately flocculant (FI between 30 and 70 %) and slightly flocculant (FI > 70 %).

For the fermentation rate, each strain was inoculated at a concentration of 6 Log CFU·mL^-1 into tubes with 15 mL of sterile must adjusted to 25 °Brix and incubated at 25 °C for eight days. Weight loss per day was measured and a negative slope was obtained (g·day^-1), in which three groups were formed according to their fermentation rate: slow (< -0.07 g·day^-1), medium (between -0.071 and -0.08 g·day^-1) and fast (> -0.081 g·day^-1), preferring those cells that presented any of the last two.

Tolerance to ethanol and SO₂ was determined by turbidimetry. First, 5 Log CFU of each strain were inoculated into individual microplate pits containing 200 μL of NYDB medium (pH = 3.8), with ethanol (8 %) and potassium metabisulfite (50 mg·L^-1 of SO₂) added. The microplates were incubated for 72 h at 25 °C and at the end the optical density was measured at 600 nm in a Variouskan^TM analyzer (Thermo Fisher®, Finland). Strains with an OD₆₀₀ > 0.7 were considered tolerant.

Pressure tolerance was evaluated from microfermentations. First, the base cider was obtained from pasteurized must and adjusted to 25 °Brix, for which the commercial yeast K1-V1116 was used. At the end of the first fermentation, 300 mL of base cider were emptied into disinfected PET bottles fitted with a Schrader valve; 30 g·L^-1 of sugar and 6 Log UFC·mL^-1 of the native yeast to be evaluated were added. Finally, the pressure was measured weekly for 49 days with a pneumatic manometer (Autotec 77992, AutoZone, Mexico).

Production of sparkling cider

‘Golden Delicious’ apple must was sulphited, racked and inoculated (1x10⁵ CFU·mL^-1) with the most outstanding native strain to obtain the base cider. The juice was fermented at 19 ± 1 °C, as proposed by ^{Soto-Herrera, Castillo-Castañeda, and Martínez-Peniche (2008)}. The cider obtained was decanted, clarified with albumin and divided into batches corresponding to the pre-selected strains to be evaluated. In each batch, tirage liquor (1x10⁵ CFU·mL^-1 of the yeast strain and 30 g·L^-1 of sucrose) and 50 mg·L^-1 of ammonium sulfate were added, and it was distributed in six 250 mL PET bottles fitted with a Schrader valve. The bottles were left to ferment at 16 ± 1 °C for 28 days. The internal pressure was measured weekly with the manometer.

At the end of the second fermentation, the sediments were removed by inverting the bottles (25 days), freezing the bottle neck and then disgorging. Finally, following the protocols recognized by the International Organisation of Vine and Wine (^{OIV, 2017}), the following were determined: alcohol by volume (ABV) by distillation and densimetry, residual sugars (RS; g·L^-1) by Fheling-Causse-Bonnans, pH by means of a potentiometer, total titratable acidity (TTA; g·L^-1) by titration, volatile acidity (VA; g·L^-1) with the García-Tena methodology and total sulphurous anhydride (SO₂) by iodimetry (mg·L^-1).

Sensory analysis

Two unstructured linear hedonic tests were carried out with limits from 0 to 10 (where 0 corresponds to “I really dislike it” and 10 to “I like it a lot”) and 10 semi-trained panelists. In the first test the foaming characteristics of the ciders were evaluated: a) crown height (foam formed on the liquid surface), b) crown reduction time, c) number of trains (corresponding to the bubble lines that form within the liquid and rise to the surface), d) bubble diameter and e) effervescence speed. The second test assessed the acceptability of appearance, aroma and taste (^{Soto-Herrera et al., 2008}).

Identification of selected yeasts

Species-level identification of the selected yeasts was performed by amplification through polymerase chain reaction (PCR) and subsequent sequencing of the ITS1/ITS4 region with the oligonucleotides ITS1 (5’ TCCGTAGGTGAACCTGCGG 3’) and ITS4 (5’ TCCTCCGCTTATTGATATGC 3’), using the following conditions: 2 min at 95 °C, 35 cycles (30 s at 95 °C, 1 min at 56.75 °C and 1 min at 72 °C) and 10 min at 72 °C. The amplicons were sent to LANBAMA of the Instituto Potosino de Investigación Científica y Tecnológica (IPICYT, SLP) for purification and subsequent sequencing with Sanger’s technique and a genetic analyzer (3130 Genetic analyzer, Thermo Fisher^®, USA). The sequences obtained were compared with those of GenBank using the Blast program.

Statistical analysis

A completely randomized unifactorial experimental design with a different number of replicates was used, with the yeast strains being the study factor. All yeast preselection determinations were performed in triplicate. The data of the quantitative variables were subjected to an analysis of variance and Tukey’s test (P ≤ 0.05), for which the JMP version 11 statistical program was used. To determine differences between the sensory characteristics of the ciders, the Friedman test was used with the Prism-GraphPad program.

Initial must characteristics

Table 1 shows the initial characteristics of each apple variety used, and it can be seen that there are important differences among them. The highest must yields were obtained with ‘429a’, ‘428, ‘Red Delicious’ and ‘Rayada’ (0.45, 0.45, 0.44 and 0.43 L·kg^-1 of apple, respectively), which contrast with ‘Malus x Micromalus’ (02.26 L·kg^-1 of apple); this is probably due to the low percentage of juice and greater firmness of the ‘Malus x Micromalus’ pulp. The highest concentration of sugars was in ‘Agua Nueva’ and ‘Royal Gala’ (17 °Brix), which differs from that obtained with ‘424’ and ‘Rayada’ (11.2 and 11.4 °Brix, respectively). Finally, the highest pH value was recorded by the must obtained with ‘428’ (pH = 4.37), and the lowest value was shown by ‘Malus x Micromalus’ (pH = 2.97).

Yeast isolation and differentiation

The evolution of spontaneous fermentations, measured by density, was very variable (Figure 1). Variety ‘436’ completed fermentation in five days, while ‘467’ and ‘Golden Delicious’ did so in 24 days. For their part, ‘Malus x Micromalus’, ‘Rayada’, ‘429b’ and ‘Joya’ did not finish the fermentation process. Many factors are involved in the development of fermentation, including the concentration of nutrients, the way in which the added sulphurous anhydride is combined, and the type of microorganisms present in the apple must, as they can compete with the fermentative yeasts for space and nutrients, or produce protein-type compounds (killer effect) or short- or medium-chain fatty acids that inhibit fermentation (^{Irastorza & Dueñas, 2010}). Rapid fermentation is generally preferred, as otherwise there is a risk of microbiological contamination and increased production costs.

Figure 1 Evolution of the must density of 14 apple varieties. 

From the spontaneous fermentations produced with the 14 varieties, 135 yeasts were isolated, as it was not possible to recover viable yeasts from variety ‘429’ at the end of fermentation. This is probably because none of the yeasts were able to tolerate the inhibitory conditions generated during the fermentation of this variety. Of the 135 isolated yeasts, 102 (75.5 %) corresponded to the genus Saccharomyces, which coincides with the findings reported by ^{Suárez-Valles et al. (2008)}, who obtained 70 % Saccharomyces spp. yeasts in isolations carried out at different stages of cider fermentation.

Characterization and selection of yeast strains

Two factors were considered in the selection of Saccharomyces spp. yeasts: β-glucosidase acitivity and the presence of the killer phenotype. Of the initial 102 strains, 60 (58.8 %) had β-glucosidase activity, which contrasts with ^{Pando-Bedriñana et al. (2010)} , who found only 2 of 134 strains. The divergence of the results may be due to the fact that in this study the yeasts were isolated from 14 varieties, separately, while in the reported work the isolates were made in Asturian cider cellars, frequently prepared with mixtures (^{Boletín Oficial de Estado [BOE], 2003}). A greater number of varieties obtained from the field increases the possibilities of obtaining phenotypic diversity from isolated microorganisms.

Glucosylated terpenes present in apples can be hydrolyzed by β-glucosidase, which releases volatile compounds that enhance the aroma and reveal typical varietal characteristics (^{Carrau, Dellacassa, & Boido, 2008}). However, the presence of the enzyme does not ensure its action in the fermentation medium, because it is associated with the yeast cell wall, which makes it susceptible to inhibition by ethanol (^{Mateo & Maicas, 2016}).

On the other hand, of the total initial strains, only 18 (17.5 %) presented killer effect. These results contrast with what was reported by ^{Pando-Bedriñana et al. (2010)}, who observed that of 134 strains isolated from cider, only two (1.9 %) exhibited this effect. These differences may be due to the origin of the yeasts. In the field, yeasts compete with a greater diversity of microorganisms, so they must develop mechanisms, such as the killer phenotype, to survive; in wineries, on the other hand, yeasts are adapted to the limiting conditions of the environment, which favor not requiring certain antagonistic mechanisms.

The secretion of toxins that inhibit the development of susceptible microorganisms acts as a mechanism for regulating population dynamics in microbial ecosystems and is considered a favorable factor for the selection of fermentative microorganisms, since they provide a competitive advantage. The susceptibility to these toxins is regulated by receptors in the cell wall of the sensitive microorganism (^{García, Esteve-Zarzozo, & Arroyo, 2016}).

In total, 66 strains showed one or two characteristics indicated above, so for the rest of the tests only these were considered (Table 2).

Regarding flocculant capacity, 30 of the 66 yeasts evaluated were moderately or highly flocculant according to the scale proposed by ^{Suárez-Valles et al. (2008)}. This characteristic is essential for yeasts, mainly in the second fermentation by the champenoise method, since to eliminate the lees in the disgorging process it is required that the yeast has the capacity to generate flocs and precipitate without adding any clarifier (^{Garofalo et al., 2016}).

In terms of fermentation rate, 38 strains showed medium or high values, which depends on the ability of each yeast to respond to stress conditions: nutrient content of the must, temperature, ethanol or fatty acid toxicity, and proportion of nitrogen sources, among others (^{Irastorza & Dueñas, 2010}; ^{Varela, Pizarro, & Agosin, 2004}). For selection purposes, if fermentation is slow, there is a risk of contamination and production costs increase as fermentation tanks are used longer; however, if it is very fast, the temperature tends to increase, which causes fermentation to stop and there is a loss of aromatic compounds. High temperature and fast fermentation are considered controllable by cooling systems, so medium to high speeds are recommended (^{Bisson & Batszuke, 2000}).

To assess the tolerance of the strains to ethanol and SO₂, only the 45 strains (including K1-V1116) that had at least two of the four previously evaluated traits (killer effect, β-glucosidase, flocculation and fermentation rate) were considered. The presence of turbidity in the medium containing 8 % ethanol and 50 mg·L^-1 of SO₂, after a 48 h incubation, indicated yeast growth and, therefore, tolerance to these two compounds; 15 strains with this trait stood out in the test (Table 2).

Ethanol is one of the most inhibitory compounds formed during fermentation, as it affects plasma membrane fluidity, vacuole morphology, glycolytic enzyme activity and mitochondrial DNA (^{Miranda et al., 2017}). ^{Mukherjee et al. (2014)} report that some yeast strains may grow or survive in the presence of ethanol due to certain alleles that counteract the toxic effects of this compound. Likewise, sterols and unsaturated fatty acids synthesized in the presence of oxygen, such as palmitoleic acid (D9-cisC16:1) and oleic acid (D9-cisC18:1), play a fundamental role in ethanol tolerance (^{García, Quintero, & López-Munguía, 2004}). In addition, it is important that the selected yeasts tolerate high concentrations of SO₂, as this will allow them to survive being sulphited, which will help them to impose themselves on wild yeasts (^{Ubeda, Briones, Izquierdo, & Palop, 1995}). The test carried out (combining two inhibitors) is more demanding and more representative than the condition presented in the fermentation medium, where various factors that limit the development of microorganisms interact.

Of the 66 yeasts evaluated, four (including reference K1-V1116) showed the five desirable traits, while 34 (51 %) presented at least three of these traits (Table 2). Of the latter, 19 (56 %) were isolated halfway through fermentation and 15 (44 %) at the end of the process, a stage in which yeasts with the best fermentative characteristics were expected to predominate, as they would be adapted to the medium and would tolerate the most inhibitory conditions present in advanced stages of fermentation (^{Steensels & Verstrepen, 2014}).

Regarding apple varieties, those with the highest number of strains with at least two desirable traits were ‘436’ (rapid fermentation), ‘428’ and ‘Royal Gala’ (with eight, seven and six strains, respectively), and with ‘467’, ‘Agua Nueva’ and ‘Rayada’ four strains were found. The varieties that contributed strains with three or fewer desirable traits were ‘Red Delicious’, ‘424’, ‘429b’, ‘Joya’, ‘Malus x Micromalus’ and ‘Golden Delicious’. In general, the varieties where fermentation was not completed, or where it took longer, were those with the fewest number of strains with positive traits; however, strain MM7, from incomplete fermentation, manifested the five desirable traits evaluated.

Considering the above results, ten strains were selected for the following tests: three that presented the five traits evaluated (MM7, 436.4 and RG6), three with four traits (AN5, RG8 and 429a.9) and four with three traits (428.3, RY3, RY5 and 428.6). In this way, different varieties and different combinations of aspects of interest were covered.

Pressure tolerance

Figure 2 shows a similar behavior in the evolution of the pressure in the bottle generated by the different yeast strains throughout the process, with the exception of strain 428.6, which generated a lower pressure than the rest of the yeasts. After 49 days, there were differences among the yeasts, especially MM7, a strain similar to the commercial control K1-V1116 and contrasting with 428.6.

Figure 2 Pressure generated by ten strains of native yeasts and commercial control K1-V1116 inside bottles. ^zMeans with the same letter do not differ statistically (Tukey, P ≤ 0.05).  

The carbon dioxide in the medium is inhibitory to yeasts, as it can slightly lower the pH, although, fundamentally, it increases atmospheric pressure, which limits the development of microorganisms. Yeasts tolerate around 7 atm (maximum reached in sparkling wines), as a higher pressure inhibits their metabolism and slows their development (^{Borrull, Poblet, & Rozès, 2015}). Like many types of stress, pressure decreases the cell cycle and reduces the viability of microorganisms as it increases; above 30 atm it generates inhibition and when it is greater than 500 atm it alters the morphology of microorganisms. Additionally, it is known that, like other stressors, pressure can generate cross-tolerance; that is, barotolerance can be acquired by sublethal pretreatment with other stressors such as temperature, ethanol or the same pressure (^{Fernandes, 2005}).

Behavior of selected native yeasts in the production of sparkling cider

Considering the above tests, strain MM7 was selected for the first fermentation, as it presented the killer effect, β-glucosidase activity and average fermentation speed. For the second fermentation, MM7 and 436.4 were chosen, as they presented the five desirable traits, and RY5 which presented three of these traits, including killer effect, average fermentation speed and tolerance to ethanol and SO₂; in addition, these three strains were the most pressure tolerant. Also, commercial yeast K1-V1116 was included.

The first fermentation took place normally, reaching a constant density by the sixth day, which indicated its end. The evolution of the second fermentation, measured by the pressure increase in the bottle, is shown in Figure 3 where it can be observed that by day seven the reference yeast K1-V1116 already had a pressure of 5.10 atm, followed by 436.4 (4.39 atm), MM7 (3.25 atm) and RY5 (2.91). Day 21 was considered the end of the second fermentation, at which time K1-V1116, 436.4 and MM7 reached a similar pressure (5.93, 5.83 and 5.75 atm, respectively), being higher than RY5 (5.49 atm).

Figure 3 Evolution of the pressure in the bottle as a function of the yeast used for the second fermentation of base cider. Means of six observations ± standard error. 

During the second fermentation, the CO₂ produced raises the internal pressure of the bottle and generates effervescence at the time of serving (^{Liger-Belair et al., 2009}). It is considered that 5 g·L^-1 of sugar increases the pressure by approximately 1 atm (^{Navarre, 1998}), which coincides with this study, because with 30 g·L^-1 of sugar, about 6 atm of pressure were obtained with the four strains evaluated. Yeast RY5 was the least efficient in this respect, which coincides with the fact that it was the least outstanding of the three in the other selection criteria.

With respect to the physical and chemical analysis of the ciders (Table 3), the three strains selected obtained values similar to those of the commercial strain. The ABV fluctuated between 8.13 and 8.89, being higher with strain RY5. The RS ranged from 1.65 to 1.96 g·L^-1, which corresponds to complete fermentations. The VA ranged from 0.04 to 0.48 g·L^-1, obtaining the highest concentration with RYS; pH, from 3.61 to 3.66; TTA, from 4.05 to 4.28 g·L^-1, and SO₂, from 202 to 226 mg·L^-1, values that are within the standards established by the Mexican official standard for alcoholic beverages (^{NOM-199-SCFI-2017, 2017}).

Sensory analysis

The Friedman test showed significant differences (P ≤ 0.05) in all sensory variables evaluated (Table 4). At crown height, cider made with 436.4 obtained the highest value in the rank sum (RS = 34.5), corresponding to a greater preference. According to ^{Esteruelas et al. (2015)}, ciders are expected to have a higher crown height at the time of serving. K1-V1116 achieved the highest values for crown reduction time (RS = 40.0), number of trains (RS = 38.5) and bubble diameter (RS = 37.5); the highest values in this last variable indicate smaller bubbles. Finally, with strains MM7 and 436.4, the highest effervescence speed values were obtained (higher speed in the ascent of the bubbles to the liquid surface) (RS = 36.0 and 34.0, respectively).

Effervescence is a key attribute in sparkling beverages, as it is the first thing consumers perceive (^{Brissonnet & Maujean, 1993}; ^{González, 2010}). The yeast used in the second fermentation of sparkling cider affects its composition, mainly due to the speed of autolysis that allows the release of polysaccharides, proteins, peptides, free amino acids, lipids and nucleotides into the medium (^{Alexandre & Guilloux-Benatier, 2008}; ^{Comuzzo et al., 2015}), which provides the cider with foam stability, decreased astringency, protection against oxidation, and greater aromatic and flavor complexity (^{Pozo-Bayón, Andujar-Ortiz, Alcaide-Hidalgo, Martín-Álvarez, & Moreno-Arribas, 2009}).

In terms of visual preference, ciders made with K1-V1116 and MM7 (RS = 35.5 and 33.5, respectively) stood out, which coincides with the results obtained in the sensory evaluation of the effervescence produced. In the olfactory aspect, the commercial strain K1-V1116 (RS = 40.0) stood out, which agrees with its technical fact sheet in which it is recognized as an ester-producing yeast that releases floral and fruity aromas (^{LALVIN, 2014}). On the other hand, with strain RY5 a very low value was obtained (RS = 10.0) in this aspect; however, this strain reached the highest preference at taste level (RS = 40.0), being above the control. This contrasts sharply with the values obtained in the rest of the variables, both sensory and analytical. In general terms, MM7 stood out among the native strains, contrasting strongly with RY5.

The results obtained show that native strains can produce sparkling ciders of quality and acceptability comparable to those produced with commercial strains. The economic benefits of having a yeast strain that produces a sparkling cider that stands out in its characteristics are that it can be recognized by the knowledgeable consumer and, therefore, can generate greater prestige and profit, as is the case with champagne.

Molecular identification of yeasts

The homologies with the database showed, with sufficient confidence (E-Values of 0 & Query Cover of 99 %), that strain RY5 belongs to Saccharomyces paradoxus (100 % ID), while MM7 and 436.4 belong to S. cerevisiae (99 and 87 % ID, respectively). Strain RY5, with the least outstanding traits, belongs to a species little related to fermentative environments; instead, its presence has been detected in the field (^{Kowallik, Miller, & Greig, 2015}). However, this strain had outstanding taste acceptance and high alcohol production, which has not been reported for cider production, so it should be further studied.