Twenty yeast strains, both Saccharomyces and non-Saccharomyces (Pichia, Kluyveromyces, Hanseniaspora, Candida, and Zygosaccharomyces), were studied (Table 1). They were isolated from different food ecosystems in Castilla-La Mancha. The non-Saccharomyces had been identified up to strain level in previous studies (2). All of them were conserved at −80°C in the culture collection of the Yeast Biotechnology Laboratory (University of Castilla-La Mancha, Spain) and, before using, a fresh culture on YPD agar (Yeast Peptone Dextrose) (Scharlau) at 30°C/24 h was obtained.

The strains were selected for this study according to their good probiotic potential in vitro, shown in previous works (2, 4). Saccharomyces boulardii CNCM I-745 was used as a positive control, since it is the only commercial probiotic yeast. All the assays were performed in triplicate.

A bacterial culture of Staphylococcus aureus (CECT 86) was used as positive control in some virulence factors.

The genetic variability of the Saccharomyces isolates and the commercial probiotic yeast was evaluated by PCR amplification (in a Perkin–Elmer GeneAmp PCR System 2400) of inter-δ region with primer pair δ2 (5′-GTGGATTTTTATTCCAAC-3′)–δ12 (5′-TCAACAATGGAATCCCAAC-3′) and δ12–δ21 (5′-CATCTTAACACCGTATATGA-3′) following the protocol described by Legras and Karst (34).

δ-PCR products were analyzed by electrophoresis in 1.5% agarose gel, according to the standard procedure for identifying at strain level.

Antibiotic resistance was carried out using Liofilchem© antibiotic strips. Erythromycin (0.016–256 mg mL⁻¹), Gentamicin (0.064–1,024 mg mL⁻¹), Chloramphenicol (0.016–256 mg mL⁻¹), Tetracycline (0.016–256 mg mL⁻¹), and Ampicillin (0.016–256 mg mL⁻¹) were used. The strips containing the antibiotics were placed onto the surfaces of YPD Agar plates inoculated with the different yeasts (10⁶ cfu mL⁻¹). The system was incubated at 30°C for 24 h. This was a modified E-test, thus the presence of a halo indicated inhibition of the target test. The shape of the halo was an ellipse and the intersection of the lower part of the ellipse with the test strip indicated the MIC value (Minimal Inhibitory Concentration).

Resistance to certain common antifungals was evaluated. The methodology used was as follows. The yeasts were seeded on Petri dishes using the microbial turf technique and then a drop of 5 μL of each antifungal solution was added. The plates were incubated at 30°C for 48 h. After this time, the halo around the drops indicated yeast inhibition.

The sensitivity of the strains against the antifungal compounds was determined by measuring the diameter (mm) of total (complete absence of growth) and partial (slight growth) inhibition zones.

The antifungals used were Nystatin (21 mg mL⁻¹) (Mycostatin®, Bristol-Myers Squibb, S.A.), Ciclopirox-Olamine (1 mg mL⁻¹) (Cyclochem®, Ferrer), Clotrimazole (10 mg mL⁻¹) (Canesten®, Bayer), and Fluconazole (150 mg mL⁻¹) (Fluconazole®, Cinfa).

Before analysis, with the aim of promoting enzyme induction, 6 passes were made from the preculture of each strain in YPD broth supplemented with 0.1% (w/v) of each precursor amino acid (L-histidine monohydrochloride, L-ornithine monohydrochloride, tyrosine disodium salt and L-lysine monohydrochloride) (Sigma, St. Louis, USA) and 0.005% (w/v) of pyridoxal-5-phosphate at 30°C for 24 h. Then the cultures (10⁶ cfu mL⁻¹) were centrifuged (15,000 × g, 5 min, 4°C) and cell-free supernatants were used for the assay. A non-inoculated medium was used as a control.

Quantification of 7 BAs (histamine, tyramine, putrescine, cadaverine, tryptamine, 2-phenylethylamine, and spermidine) was carried out by RP-HPLC using a diethyl ethoxymethylenemalonate (DEEMM) derivatization method (35). The samples stored at −20°C were thawed and immediately derivatized. For that purpose, a mixture of 1 mL of the sample, 1.75 mL of 1 mol/L borate buffer pH 9.0, 750 μL of methanol and 30 μL of DEEMM was incubated at 30°C in an ultrasonic bath for 30 min. The samples were then heated at 70°C for 2 h to allow the complete degradation of excess DEEMM and reagent by-products. After derivatization, the samples were filtered through regenerated cellulose esters 0.2 μm membranes (Análisis Vínicos, Tomelloso, Spain) coupled to a syringe into conical vials.

The analysis was performed using an Agilent 1200 HPLC (Agilent Technologies, Madrid, Spain) equipped with a Zorbax Eclipse XDB C18 column particle size 5 mm (250 mm × 4.6 mm), an Agilent guard cartridge C18 particle size 5 mm (12.5 mm × 4.6 mm) and a photodiode array detector (Agilent Technologies, Madrid, Spain). The mobile phase consisted of a 50 mmol L⁻¹ acetate buffer pH 8.75 with 0.02% sodium azide (eluent A) and an 80:20 mixture of acetonitrile and methanol (eluent B). Samples (50 μL) were applied to the column and eluted at a flow rate of 0.9 mL min⁻¹ according to the binary gradient shown in Table 2. The target compounds were identified by their retention times and their spectral characteristics at 280 nm and were quantified using the external standard method. The limits of quantification for the 7 amines were: histamine: 0.13 mg L⁻¹; tyramine: 0.07 mg L⁻¹; putrescine: 0.03 mg L⁻¹; cadaverine: 0.03 mg L⁻¹; tryptamine: 0.13 mg L⁻¹, 2-phenylethylamine: 0.07 mg L⁻¹, and spermidine: 0.20 mg L⁻¹.

The BSH activity was determined as described by Du Toit et al. (36), with slight modifications. Fresh yeast cultures (10⁶ cells mL⁻¹) were streaked in triplicate on YPD agar containing 0.5% (w/v) taurodeoxycholic acid sodium salt (TDCA; Sigma, USA) and 0.037% (w/v) CaCl₂. The plates were incubated at 37°C for 72 h.

For qualitative assaying of the deconjugation activity, the formation of an opaque or whitish halo zone around the biomass due to the release of free bile acids was observed (37).

A one-way analysis of variance (ANOVA) was applied to the BA results as well as comparing the Pz values of yeast isolates in the secretion of the hydrolytic enzymes studied, using the Student-Newman-Keuls (S-N-K) test for comparison of the means (P < 0.05). All the statistical analyses were performed using the IBM SPSS statistics package ver. 24.0 (SPSS Inc., Chicago, IL, USA).

As can be seen in Figure 2A, when amplifying the inter-δ region with the primer pair δ2-12, strains 7, 95 and 127 were apparently the same, but when using the primers δ12–δ21, the genetic profiles were different. The same occurred with strains 3, 127, 128, 132, and 139 whose bands obtained with the primers δ12–δ21 (Figure 2B) were very similar, but with the primer pair δ2–δ12 it was observed they were not the same strains. Therefore, combining both results, it was demonstrated that the 10 S. cerevisiae of the present study (isolated from different wineries) are distinct from each other and that they also have different profiles from the strain marketed as the only probiotic yeast (26).

All yeasts evaluated showed resistance to the antibiotics studied. The same data are shown by other authors (42). On the contrary, none of the 20 yeasts showed resistance to any of the antifungal products used in this study (Nystatin, Ciclopirox-Olamine, Clotrimazole, and Fluconazole) except H. osmophila (1056) which, in contact with Ciclopirox-Olamine, did not show any halo of inhibition.

The results of the inhibition halo tests with the antifungal compounds are shown in Figure 3. In the case of Nystatin and Ciclopirox-Olamine, the strains with the highest resistance, and therefore the smallest halo of inhibition, were non-Saccharomyces, unlike what occurred with Clotrimazole and Fluconazole, for which the Saccharomyces strains were the most resistant. Nystatin and Clotrimazole were able to totally inhibit the growth of all the strains, although the latter was the most effective. In contrast, Fluconazole showed the lowest effect and only inhibited a total of 8 strains, with the rest showing a partial effect. In all cases, the strains showed significant (P < 0.05) differences between them. On the other hand, taking into account the global antifungal effect, it was observed that the most resistant strains were S. cerevisiae 3 and H. osmophila 1094, while the most sensitive strain was L. thermotolerans 1039 (Figure 4 shows a clear example of total and partial inhibition).

These results agree with those reported by Abulreesh et al. (43). Maciel et al. (44) also observed how different non-Saccharomyces yeasts were susceptible to antifungals, including those of the Pichia genus which was susceptible to Fluconazole.

Seventeen of the 20 yeast strains evaluated in this study can be considered as non-amine or very low-amine producers (<1 mg L⁻¹, Table 3). Among the four remaining strains, P. kudriavzevii 1200, P. anomala 1090, and Z. bailii 1213 produced mainly histamine in concentrations below 23 mg L⁻¹ and with significant (P < 0.05) differences between them. On the other hand, the H. osmophila 1094 strain accounted for the highest total BA value, mainly due to a very high production of tyramine (almost 700 mg L⁻¹) and also to a considerable amount of 2-phenylethylamine (nearly 65 mg L⁻¹). It was the only strain which produced these two amines and it also produced histamine in a lower quantity (8.45 mg L⁻¹). It is considered that, in terms of good fermentation practices and safe consumption, a fermented food should not exceed 200 mg kg⁻¹ of total BAs (45). Among the strains evaluated, all the yeasts except H. osmophila 1094 meet this criterion (they produce <200 mg Kg⁻¹ of total amines).

These results show that BA production was proven as a strain-specific property, as has been previously reported for bacterial strains (46, 47). The production of histamine has been reported for Candida and Pichia strains in concentrations below 5 mg L⁻¹ (48). Kung et al. (31) found 6 histamine-producing yeast strains belonging to the Candida glabrata and Candida rugosa species, capable of producing from 9.2 to 41.7 mg L⁻¹ of histamine. Chang et al. (32) reported a strain of Zygoascus hellenicus var. hellenicus capable of producing amounts of histamine (14.6 mg L⁻¹) comparable to those obtained in this study. On the other hand, the production of tyramine by strains of Kluyveromyces lactis (205 mg L⁻¹) and Yarrowia lipolytica (198 mg L⁻¹) has been reported (48), these values being much lower than those produced by strain 1094 of Hanseniaspora osmophila in the present work. The production of 2-phenylethylamine by yeasts has been reported for the different yeast strains of Kloeckera, Metschnikowia, Brettanomyces, Candida, Rhodotorula, and Trichosporon (49, 50). However, in all cases, the concentrations were much lower than those produced by the H. osmophila 1094 strain in the present work.

The results of the virulence factors studied are shown in Table 4. It was observed that 55% of the yeasts showed BSH activity. The 4 strains of the Pichia genus (1003, 1200, 1082, and 1090) were positive, unlike those corresponding to Hanseniaspora (1056 and 1094). Regarding S. cerevisiae, not all the strains presented the same trend, with some of them being positive and others being negative. The results obtained were also unrelated to their origin. The commercial probiotic turned out not to show any activity.

The results were not related between genders, nor were correlations found with other studies. Tanaka et al. (51) already reported that the molecular weight and structure of BSH is strain dependent. Istiqomah et al. (52) found BSH activity in 104 yeasts (from among 112 isolated and tested) belonging to Saccharomyces, Candida, and Cryptococcus, while others like Sourabh et al. (53) found no activity in any of the 23 yeast isolates tested which had been obtained from various traditional fermented foods of the Western Himalayas.

Deconjugation has been included by the World Health Organization (WHO) experts as one of the main activities of intestinal microorganisms (14) since probiotics could maximize their survival prospects in the hostile environment of the gastrointestinal tract. In addition, this activity has been correlated with the reduction of the cholesterol level in the blood (23). However, it is not yet completely clear whether BSH activity is a desirable property, since large amounts of deconjugated bile salts may have undesirable effects for the human host (54).

The coagulation test was performed on citrated rabbit plasma and showed that all yeasts gave negative results, being unable to coagulate the rabbit plasma at the incubation times tested (2, 4, 6, and 24 h) at 37°C (Table 4), unlike S. aureus CECT 86 (positive control) which was able to coagulate it in the first hour.

This shows the difference from yeast pathogenic strains, such as Candida albicans, which do show positive activity (55, 56) like other Candida species isolated from patients with a pathology (56). No data was found in the literature on coagulase activity in any other strains except those of the Candida genus.

Deoxyribonuclease activity (DNAse) was studied in two different culture media. In DNAse agar medium with methyl green, all strains belonging to the Saccharomyces genus were able to grow but none showed enzymatic activity. Non-Saccharomcyes yeasts did not grow in this media. In DNase agar without methyl green, all the yeasts grew although none of them showed hydrolysis halos with the addition of hydrochloric acid (Table 4). Only the S. aureus strain, used as a positive control, could grow and showed a hydrolysis halo in both media, although it was more intense in the medium without methyl green.

Although there is little information about DNAse activity in yeasts, the strains with this activity have been isolated from clinical environments and immunosuppressed patients, for example Cryptococcus neoformans and Cryptococcus gattii (15) or C. albicans (16, 38). Investigation of DNAse activity in yeast isolates requires standardization and technique perfection in order to improve sensitivity in the detection of the activity and mechanisms of action of this enzyme (40).

None of the 20 strains studied showed this activity in TSA-blood medium after 48 h although all of them were able to grow.

Hemolysin expression is believed to be an inherent factor that is triggered under specific conditions (57). In yeasts, hemolytic factor is known to be a putative virulence factor contributing to pathogenicity in the Candida species (57). Furthermore, only certain strains of this genus, isolated from patients with pathologies, have been reported to produce this activity (58, 59).

Extracellular secretion of proteases and phospholipases were analyzed by measuring the precipitation area around each colony after growing on BSA and Egg-Yolk media, respectively. The qualitative results and Pz values are shown in Table 4 and Figure 5.

All yeasts were able to grow on BSA agar and showed high proteolytic activity (Pz ≤ 0.50) with significant (P < 0.05) statistical differences (Figure 5A). Strain 127 showed the lowest value (0.47), with significant (P < 0.05) differences from the rest. On the other hand, strains 7, 95, and 1200 showed the highest (P < 0.05) values along with the commercial probiotic yeast (strain 24). The average Pz value for this activity was 0.38.

Moradi et al. (60) also found proteolytic activity in Saccharomyces sp. and Kluyveromyces sp. (isolated from sweet fruit and dairy samples). Ombrella et al. (61) detected that more than 60% of the C. albicans studied showed a high degree of proteolytic activity. De Llanos et al. (20) and Llopis et al. (21) showed that most Saccharomyces isolates from dietary supplements also present this activity. In the case of Jafari et al. (17), who evaluated pathogenic strains, high proteolytic activity was observed in addition to coagulase activity, unlike the strains studied here that did not present this last activity.

The results for phospholipase activity showed that all the studied strains could grow in this medium but 9 of them showed no activity (Pz = 0.50). Therefore, according to the S-N-K test from ANOVA, two groups were clearly differentiated: those with a Pz-value of 0.43–0.48 and on the other hand, those with a Pz value of 1 (Figure 5B).

De Llanos et al. (20) carried out a comparative study between clinical and industrial S. cerevisiae yeast strains based on the various phenotypic traits associated with pathogenicity. They showed that, like the pathogenic strains, those isolated from food sources as well as S. boulardii secreted high levels of protease and phospholipase, which coincides with the results obtained in the present study. Therefore, neither activity could be considered a relevant aspect in determining whether yeasts are safe or not. Similarly, Llopis et al. (21) found yeasts included in dietary supplements with phospholipase activity levels ranging from low to moderate. S. boulardii was included in the low activity group, unlike in the present work where it showed no activity.