Pathogenic Potential of Saccharomyces Strains Isolated from Dietary Supplements

Saccharomyces cerevisiae plays a beneficial role in health because of its intrinsic nutritional value and bio-functional properties, which is why it is also used as a dietary supplement. However, the perception that S. cerevisiae is harmless has changed due to an increasing number of infections caused by this yeast. Given this scenario, we have tested whether viable strains contained in dietary supplements displayed virulence-associated phenotypic traits that could contribute to virulence in humans. We have also performed an in vivo study of the pathogenic potential of these strains using a murine model of systemic infection by intravenous inoculation. A total of 5 strains were isolated from 22 commercial products and tested. Results highlight one strain (D14) in terms of burden levels in brains and kidneys and ability to cause death, whereas the other two strains (D2 and D4) were considered of low virulence. Our results suggest a strong relationship between some of the virulence-associated phenotypic traits (ability to grow at 39°C and pseudohyphal growth) and the in vivo virulence in a mouse model of intravenous inoculation for isolates under study. The isolate displaying greatest virulence (D14) was evaluated in an experimental murine model of gastrointestinal infection with immunosuppression and disruption of mucosal integrity, which are common risk factors for developing infection in humans, and results were compared with an avirulent strain (D23). We showed that D14 was able to spread to mesenteric nodes and distant organs under these conditions. Given the widespread consumption of dietary supplements, we recommend only safe strains be used.


Introduction
The yeast Saccharomyces cerevisiae is well known mainly for its use in the production of foods (bread) and of alcoholic beverages (wines, beers, ciders, saké) through alcoholic fermentation. In addition to its use in fermentative processes, S. cerevisiae is outstanding for its intrinsic nutritional value, a reason why it is also used in dietary supplements. Moreover, these preparations are believed to possess some bio-functional properties that exert positive effects on one or more of the host's physiological functions due to its content in minerals and B-complex vitamins. The possible beneficial effects of consumption are well known [1] and there are studies into its medicinal properties [2,3,4,5]. This is why the consumption of nutritional yeasts is a current trend promoted by healthy diets and lifestyles.
Although the bio-functional properties do not depend on whether yeast cells are alive (the majority of commercial products contain deactivated cells), some companies include revivifying cells in their products, as indicated on the labels. In these cases, living cells enter the body in ongoing and high concentrations (an average daily intake of 2.1610 7 cells is recommended). Despite this growing trend, there has not been a great deal of interest in the possible undesirable effects of consumption. This is mainly because S. cerevisiae has always been considered a safe micro-organism for nutritional use (''GRAS'', Generally Regarded as Safe) without considering its undesirable facets. However, this concept is currently changing due to the increasing incidence of infections associated with this yeast species; a comprehensive review of the topic shows that from 92 documented cases of invasive Saccharomyces infection, 15 were diagnosed before 1990, while 76 cases were diagnosed after 1990 [6].
Saccharomyces cerevisiae fungemia in humans mainly occurs in immunodepressed patients. Nevertheless, the existence of strains with inherent virulent potential cannot be ruled out since, although it is not common, cases of fungemia by this yeast have also been described in healthy hosts [7,8,9]. The few studies assessing the pathogenic potential of this yeast species are indicative of the aforementioned risk, and show that most clinical isolates of S. cerevisiae exhibit certain phenotypic characteristics [10,11,12,13,14] and display different degrees of virulence when tested in murine models [15,16,17]. These phenotypic traits include growth at 42uC, some hydrolytic activities, pseudohyphal and invasive growth, switching and adhesion, which have been associated with virulence in pathogenic microorganisms [18][19][20][21][22][23][24][25][26][27]. Some of these traits have been studied more recently in our laboratory in a large number of clinical and non-clinical isolates [28]. Activation of signal transduction pathways mediated by the Mitogen-Activated Protein Kinases (MAPK) Kss1, which regulates pseudohyphal and invasive growth, and Slt2 from the cell wall integrity pathway (CWI) [29], have been also analyzed in these strains [30]. Interestingly, we observed that a commercial baker's yeast as well as S. cerevisiae var. boulardii (strain isolated from a commercial lot of the bio-therapeutic agent Ultra-Levura) appeared to be related to clinical strains based on such phenotypic traits [28], and they presented remarkable dissemination capacity in murine models of systemic infection [31]. Furthermore, they were encountered in the blood of individuals with symptoms of infection (baker's yeast [32], S. cerevisiae var. boulardii [33,34,35]). Given these considerations one might ask whether the ingestion of live S. cerevisiae cells, through the consumption of these products, can cause infection in humans. Indeed, a case of fever of unknown origin was subsequently associated with the prolonged ingestion of a dietary supplement containing between 10 7 and 10 8 of S. cerevisiae per gram of product [36], which suggests that such doubts may be founded. In addition, previous results obtained in our laboratory showed that virulence in two murine models of a strain (named D14) from a commercial dietary supplement was notable compared with a group of clinical isolates [37].
Since dietary supplements are a route of entry by which live S. cerevisiae cells enter the human body, and given there are no safety studies to date, we undertook an in-depth study of the virulence of the aforementioned isolate D14, and other isolates from commercial products. We first identified the isolates molecularly to verify their adscription to the S. cerevisiae species since misidentified strains are found quite frequently [38]. Moreover, we studied whether hybrid strains between S. cerevisiae and other closely related Saccharomyces species are present in yeast-enriched foods and dietary products. In fermentative wine processes, several wine Saccharomyces strains considered to be S. cerevisiae have been reported to be hybrids [39,40]. Hybrid generation could provide adaptive values [41] hence it is interesting to evaluate whether this affects the development of infection in mammalian hosts. We analyzed whether the isolates exhibited phenotypic traits (in particular, growth at high temperatures, secretion of hydrolytic enzymes, pseudohyphal and invasive growth, colony phenotype switching, MAPK activation and plastic adherence) associated with virulence in other pathogenic yeasts before evaluating their virulence in vivo. We then studied the potentially virulent strains (those with the highest percentage of positive traits) in a murine model. To do so, we evaluated death rates and determined Table 1. Commercial products analyzed in this study.  burdens in brains and kidneys in immunocompetent BALB/c after intravenous inoculation and compared them with an isolate displaying fewer positive traits. Finally, and taking into account that the natural portal of entry of live yeast cells is oral intake, we tested a murine model of disseminated infection derived from translocation across the gastrointestinal mucosa to determine whether the strains with invasive infection ability pose a real potential safety risk. Given the complexity of such murine oral inoculation models, we tested only the isolates showing the highest and the lowest virulence.

Strains
Yeast strains used for the elaboration of commercial dietary products were isolated from 22 products (D1-D4 and D6-D23), listed in Table 1. These products included two enriched beers and 20 dietary products. S. cerevisiae strains used as control in the in vitro study of virulence-associated phenotypes and in the in vivo virulence study by intravenous inoculation (i.v.) in mice are listed in Table 2. The Candida albicans SC5314 strain [42] was used as positive control in the assay of adherence to plastics.
Then, 50 ml of each culture were spread onto YPD plates [YPD medium+2% w/v agar (Panreac, Barcelona, Spain)] and incubated 24-48 h at 30uC. A total of six colonies, from each of the products that exhibited yeast growth, were randomly chosen for further identification and characterization.
For infection experiments Saccharomyces strains were grown at 30uC on YPD plates. After 24 h, cells were harvested, washed twice with sterile phosphate-buffered saline (PBS) and diluted to the desired concentration in the same buffer for intravenous injection or oral administration.

PCR Amplification and Enzymatic Restriction
DNA was isolated according to Querol et al. [43] with the modifications described by de Llanos et al. [44].
For PCR amplification, DNA was diluted to 1-50 ng/ml. The amplification of the ribosomal 5.8S-ITS region was carried out with primers its1 and its4 [45] under the reaction and amplification conditions described previously by de Llanos et al. [44]. PCR products (15 ml) were digested without further purification with the restriction endonucleases AccI, CfoI, HaeIII and HinfI (Boehringer Mannhein, Germany). Restriction pattern comparisons were performed using the Yeast-id database (http:// www.yeast-id.org).
Saccharomyces interspecific hybrids were identified by PCR amplification of three nuclear genes APM3, PEX2 and BRE5 located in different chromosomes (II, X and XIV, respectively) and their subsequent RFLP using the endonucleases MspI, HaeIII and HinfI (Boehringer Mannhein, Germany) respectively. PCR amplification conditions and pair of primers were those described previously by González et al. [39]. Characterization was performed by comparing the restriction patterns obtained with those described in previous works for type reference strains of the species Saccharomyces uvarum, S. cerevisiae and S. kudriavzevii [39,46]. d sequences were amplified using primers d1 and d2 as previously described by Ness et al. [47] and amplification conditions were according to de Llanos et al. [44]. PCR products and restriction fragments were visualized on 1.4% and 3% agarose (Pronadisa; Laboratorios Conda S.A., Madrid, Spain) gels, respectively. A 100-bp DNA ladder marker (Invitrogen Life Technologies) served as a size standard.
For mtDNA restriction analysis, DNA was resuspended in 13 ml of water and used in full in the reaction. The restriction analysis was performed according to Querol et al. [43]. DNA was digested with HinfI (Boehringer Mannhein, Germany), following the supplier's instructions. Restriction fragments were visualized on 0.8% agarose (Pronadisa; Laboratorios Conda S.A., Madrid, Spain) gels. The DNA of phage l digested with PstI served as size standard.

Virulence-associated Phenotypes of Yeasts Isolated from Commercial Products
Growth at different temperatures. Yeast strains were cultured in YPD liquid medium at 30uC overnight. Yeast suspensions were adjusted to 0.5 OD at 600 nm, corresponding to 5610 6 cells/ml; serial decimal dilutions were performed and 5 ml of them were spotted (in triplicate) on YPD plates and incubated at 30, 37, 39 or 42uC for 1-3 days.
To determine generation time, strains were cultured in liquid YPD at 30uC and 37uC. Absorbance of the cultures was measured at 600 nm every hour for 8 h. Linear regression was used to calculate specific growth rate and generation time. The statistic t-Test was performed on these data.
Secretion of hydrolytic enzymes (protease and phospholipase). Protease production was determined according to Aoki et al. [48]. The test medium consisted of agar plates containing bovine serum albumin (BSA): 60 ml of solution was prepared containing 0.04 g MgSO 4 . 7H 2 O (Panreac, Barcelona, Spain), 0.5 g K 2 HPO 4 (Panreac, Barcelona, Spain), 1 g NaCl (Panreac, Barcelona, Spain), 0.2 g dried yeast extract (Pronadisa, Madrid, Spain), 4 g glucose (Panreac, Barcelona, Spain) and 0.5 g BSA (Fraction V, Sigma, USA), the pH was adjusted to 5 with 1N HCl. The solution was sterilized by filtration and mixed with 140 ml of melted agar, 20 ml of this medium was poured into each Petri dish and 5 ml of a 30uC overnight yeast culture adjusted to 0.3-0.4 OD at 600 nm were inoculated in each Petri dish and incubated at 37uC for 4 days.
In both assays, activity was visualized as an area of precipitation around each colony. The value of enzymatic activity (Pz) was measured according to the method of Price et al. [49] and was expressed as the ratio of the diameter of the colony alone to the diameter of the colony plus the precipitation zone; thus a low Pz indicated high enzyme production. The average Pz value was obtained using three separate samples of each strain.
To detect invasiveness, strains were patched carefully onto YPD plates to avoid scratching the agar, incubated at 30uC for three days, and then at room temperature for two additional days. A gentle stream of deionized water was then used to rinse all of the cells from the agar surface in order to observe the presence of cells growing below the agar surface.
Colony phenotype switching. Yeast strains were cultured in YPD liquid medium at 30uC overnight. Cells were collected by centrifugation and resuspended in sterile water to a concentration of 10 3 cells/ml, and 100 ml of this suspension was spread onto YPD agar plates with 1% phloxine B (Sigma-Aldrich) [12]. Plates were incubated for five days at 30uC and colonial phenotypes were then observed and counted. Stability and reversion were studied for the most abundant secondary phenotypes by the assay described above.
Adherence to polystyrene. To determine adherence to polystyrene Petri plates, yeast strains were cultured in liquid YPD at 30uC overnight. Cells were collected by centrifugation and resuspended in glucosaline solution (0.2 M D-glucose, 0.05 M NaCl) to reach a 10 6 cells/ml concentration. Ten ml of each cell suspension were added to polystyrene Petri plates of 90 mm of diameter (J. D. Catalan, S.L., Spain) and to glass tubes as control of no plastic adhesion, and incubated for 1 h at 37uC with 5% CO 2 . Supernatants containing non-adhered yeast cells were collected. Plates were washed twice with 5 ml of sterile water and washings were added to the collected suspension. These yeast cell suspensions were plated on YPD and incubated for two days at 30uC. Colonies were then counted and the difference between plate and glass tube counts was considered as the number of adherent cells.
To measure adherence to polystyrene 96-well microtiter plates, we used the protocol described by Peeters et al. [53] with some modifications. Yeast strains were cultured in liquid YPD at 30uC overnight. Cells were collected by centrifugation and resuspended in glucosaline solution (0.2 M Glucose, 0.05 M NaCl) to reach a 10 8 cells/ml concentration. One-hundred ml of each suspension were inoculated into 16 wells of a 96-well polystyrene cell culture microplate (Greiner Bio-one) and 16 wells were filled with sterile glucosaline solution as negative controls. Microtiter plates were incubated for 1 h at 37uC with 5% CO 2 . Supernatant containing non-adhered cells was removed, wells were rinsed with glucosaline solution, and cells adhering to the wells were fixed with 99% methanol (15 min) and stained with 0.02% crystal violet solution (Panreac, Barcelona, Spain) for 20 min. Plates were washed to remove the excess of crystal violet. Finally, crystal violet from stained adhered yeast cells was extracted with 33% acetic acid, and absorbance measured at 595 nm.
Adherence to polyurethane intravenous catheters. Multilumen central venous catheterization sets (Arrow International) were used. Catheters were cut into 1 cm-long sections under sterile conditions. Yeast strains were cultured in liquid YPD at 30uC overnight. Cells were collected by centrifugation and suspensions containing 10 7 cells/ml on either glucosaline solution or RPMI-1640 medium (Lonza) were obtained. One ml of each suspension was added to an eppendorf tube containing a catheter section. Eppendorf tubes were incubated for 1 and 24 h at 37uC in a rotating device. Catheter sections were then placed in new tubes and washed five times with sterile water. In each washing, tubes were gently inverted 10 times and water was removed by a glass Pasteur pipette connected to a vacuum pump. Then, each catheter section was transferred to a tube and 1 ml of sterile water was added. Each tube was shaken using a Vortex mixer for 10 minutes. Yeast suspensions were spread onto YPD plates and incubated two days at 30uC and then, colonies were counted.

Mouse Model of Systemic Infection by Intravenous Inoculation
Animals. Two to four batches of 10 mice belonging to cosanguineous strain BALB/c (immunocompetent) were used per Saccharomyces strain. All animals were obtained from Harlan Ibérica (Barcelona, Spain) and were 6-week-old female mice. On arrival, animals were left for between 4 to 5 days to acclimatize and provided water and food ad libitum.
Animal studies were carried out at the animal unit of the Central Support Service to Experimental Research of the University of Valencia (School of Pharmacy, University of Valencia, Spain). Animal manipulation was carried out in compliance with Spanish laws and guidelines (RD1201/2005 and law 32/2007) and according to the European Directive 2010/ 63/EU concerning the protection of animals used for scientific purposes. All the assays were approved by the Consejo Superior de Investigaciones Científicas (CSIC, Spain) Bioethics Committee.
Experimental infections. Mice were infected by intravenous (i.v.) inoculation into the tail vein with 2610 7 viable CFU of yeast in a volume of 0.2 ml of PBS, as previously described [16] and were sacrificed by cervical dislocation to determine yeast burdens in brains and kidneys at different time points as follows. On day +7, between two and four mice per batch were sacrificed except when death was detected prior day 7, in which case only one mouse was sacrificed reserving the surviving mice to determine burdens at days +15 and +30. The number of mice sacrificed per strain at these time points varied from two to four per batch. A control mouse was intravenously inoculated with 0.2 ml of sterile PBS alone. Organs were processed as described below.

Mouse Model of Systemic Infection by Oral Inoculation
Animals. Twelve mice belonging to strain ICR/Swiss (outbread immuno-competent) were used per Saccharomyces strain. All animals were obtained from Harlan Ibérica (Barcelona, Spain) and were 5-6-week-old female mice. On arrival, animals were left for 10 days to acclimatize and provided water and food ad libitum. Animals were caged using sterilized bedding and provided sterilized food and water from the first day of antibiotic administration.
Animal studies were carried out at the University of Valencia mentioned above.
Experimental infection. To deplete the indigenous gastrointestinal microbiota and facilitate the presence of yeast in the gastrointestinal tract, mice were fed for five days prior to oral inoculation with sterile water containing three antibiotics as described for Candida glabrata [54]: 1 mg/ml bacitracin (Sigma-Aldrich Química, S.A. Madrid, Spain), 2 mg/ml streptomycin sulfate (Reig Jofré Laboratories, S.A. Barcelona) and 0.1 mg/ml gentamicin sulfate (Normon Laboratories, S.A. Madrid). The antibiotics were administered until the end of the experiment to reduce secondary infections. To assess antibiotic effectiveness, stools were collected during three consecutive days prior to oral inoculation (at days 23, 22 and 21) and at days 1, 3, 5 and 7.
A single dose of 0.2 ml of an inoculum containing 1610 9 of live yeast cells was administered to mice by intragastric intubation (day 0). In addition, the same strain was added to sterile water with the aforementioned antibiotics at approximately 7610 9 cfu/ml and was administered via the drinking water to mice for three consecutive days (days 0-2). Then, the inoculum was removed and replaced by drinking water with antibiotics for the remainder of the experiment.
Immunosuppression. Three days after the beginning of yeast administration, mice were given the immunosuppressant drug dexamethasone (Fortecortín 40 mg, Merck Farma y Química, S.L. Madrid, Spain) at 0.1 mg/g by intraperitoneally (ip) injection twice a day during four consecutive days (days 3-6). This drug was chosen because of its widespread clinical use and because it facilitates the extra-intestinal dissemination of C. glabrata despite its relatively low pathogenicity compared to C. albicans [54]. Non inoculated animals receiving antibiotics and immunosuppression regimen were used as internal control to assess the incidence or deaths resulting from the effect of treatments.

Evaluation
of yeast translocation and dissemination. Four mice per Saccharomyces strain tested were sacrificed by cervical dislocation 5, 6 and 7 days after oral inoculation. Peyer's patches, mesenteric lymph nodes (MLNs), livers, kidneys and brains were removed for quantitative analysis of yeast burden in each organ as described below.

Determination of Burdens in Representative Organs and Stools
When required, stools were obtained by manually pressing the lower abdomen of mice, weighed and placed in 5 ml volumes of sterile saline solution. Mice were sacrificed by cervical dislocation and brains and kidneys or brains, kidneys and livers (depending on the assay: i.v. or oral inoculation) were obtained aseptically with forceps and scissors. In the studies of gastrointestinal infection, the intestine was then slowly draw out of peritoneal cavity with forceps and exposed on sterile gauze. The MLNs were identified and cut off. The small intestine was cut below the stomach and was first washed through with PBS using a syringe to expel the fecal matter and then thoroughly wash with PBS containing antibiotics [100 units/ml of penicillin (Sigma-Aldrich Química, S.A. Madrid, Spain), 100 mg/ml of streptomycin sulfate (Reig Jofré laboratories, S.A. Barcelona, Spain)] and an antifungal [2.5 mg/ml of amphotericin-B (Sigma-Aldrich Química, S.A.)]. Macroscopic Peyer's patches were identified and carefully dissected. All organs were kept in sterile saline solution immediately after their isolation and were later moved to a new tube with 5 ml volumes of the same solution. The organs were then homogenized with a Polytron PT 3100 instrument equipped with a 10 mm probe. Between each homogenization, the probe was cleaned and sterilized by two successive washes in non-sterile water, two successive washes with 70% ethanol and a final wash in sterile water. The number of viable yeast cells in the organs was determined by counting colonyforming units (CFU), following the plate dilution method using YPD plates with chloramphenicol (10 mg/ml) after 48 h of incubation at 30uC. All burdens have been expressed as the log 10 geometric mean number of CFU per entire organ and per gram for stools.
Three colonies were selected randomly from each YPD plate containing each homogenized organ or stools, and were analyzed by mtDNA restriction analysis in order to confirm they belonged to the Saccharomyces strain inoculated.

Statistical Analysis
To compare among strains based on the study of virulenceassociated phenotypes and to analyze the association between these phenotypic traits and in vivo virulence by intravenous route, a presence and absence matrix was prepared from which the Spearman's Rho correlation coefficient was calculated both for strains and traits. Moreover, Jaccard proximities were calculated excluding joint absences from consideration. P values ,0.05 were considered statistically significant (2-tailed).
In the in vivo studies after intravenous inoculation of Saccharomyces strains, burdens between brains and kidneys were compared statistically by a two-way repeated measures ANOVA (2 betweenfactors: strains and days; 1 within-factor: organ). Furthermore, since the Levene's Test showed that variances were not homogenous in all cases, we performed several one-way ANOVA using the Tamhane's Post Hoc test for multiple mean comparisons: two analyses were performed, one for each organ (brain and kidney), to study strain effect, and two more analyses were run to study the effect of time. In the in vivo virulence studies of oral inoculation in mice, a one-way ANOVA was used to compare burdens of isolates D14 and D23 for each organ (MLN, Peyer's patch, liver, kidney and brain). P values ,0.05 were considered statistically significant for all comparisons. All the statistical analyses were performed using IBM SPSS (Statistical Package for the Social Sciences) Statistics software v21 for Windows.
The Kaplan-Meier method and Wilcoxon and log-rank test were performed to compare survival curves using GraphPad software. The level of significance between curves was set at p, 0.05.

Identification of Yeasts Isolated from Commercial Products
A total of 22 commercial products including enriched beverages and dietary products, as well as the bio-therapeutic agent sold as Ultra-Levura were microbiologically studied (see Table 1 for origin and presentation of the products). Yeast cells were isolated from only eight of them (D2, D3, D4, D6, D7, D8, D14, D23) and from Ultra-Levura (D5).
A molecular study, based on restriction analysis of the 5,8S rRNA gene and the two flanking internal transcribed sequences (ITS), was performed in six colonies randomly selected from each of the nine products to confirm their adscription to the Saccharomyces genus. All strains showed the typical 850 bp 5.8S-ITS PCR amplification fragment previously described for species of the Saccharomyces sensu stricto complex [55]. The sizes of the restriction fragments obtained with endonucleases CfoI, HaeIII and HinfI (375+325+150; 325+230+170+125; 375+365+110, respectively) suggested that, as indicated on the labels, these strains belong to S. cerevisiae species, with the exception of D6 strain. The 5.8S-ITS fragment amplified from the latter strain showed CfoI and HinfI restriction patterns typical of S. cerevisiae, but an atypical profile was obtained when HaeIII was used, specifically 500+230+ 125. According to published data, this profile may correspond to two species of the Saccharomyces sensu stricto complex (Saccharomyces bayanus and Saccharomyces kudriavzevii). Therefore restriction enzyme AccI was used as it is able to distinguish between the two species [40]. The restriction pattern obtained (730+110) suggested that isolate D6 corresponds to the species S. kudriavzevii [40].
Additional restriction analyses of the nuclear genes APM3, BRE5 and PEX2 revealed that two of the isolates were in fact hybrids between S. cerevisiae and other Saccharomyces species very closely related (species belonging to the Saccharomyces sensu stricto complex). In the case of D6, initially identified as S. kudriavzevii, the restriction patterns obtained from the digestion of the AMP3 gene with MspI (750+520+310+200), digestion of the PEX2 gene with HaeIII (345+260+110) and digestion of the BRE5 gene with HinfI (330+289+228+153) indicated the presence of a hybrid between S. kudriavzevii and S. cerevisiae. The other isolate showing different adscription was D23; the restriction analysis of the nuclear genes AMP3 with MspI (750+680+340+280), PEX2 with HaeIII (345+ 260+180+150+110) and BRE5 with HinfI (289+228+100+66+60+ 43) revealed a hybrid between S. cerevisiae and S. bayanus species.

Molecular Characterization of Yeasts Isolated from Commercial Products
We studied the purity of the nine commercial products containing live yeasts by the molecular technique based on mtDNA restriction analysis with the endonuclease HinfI. All six randomly selected colonies isolated from the same product showed the same restriction pattern in all cases. Moreover, strains were characterized molecularly to avoid the use of repeated strains in the in vitro and in vivo pathogenicity studies. Inter-product comparisons gave a total of five different mtDNA restriction patterns ( Figure 1A). Results are summarized in Table 3 by assigning a distinct letter to each pattern differing from the others. Four restriction patterns (groups B m , C m , D m , E m ) contained only one strain indicating that these strains differed from each other. One of these restriction profiles, namely B m , corresponded to D5 (Ultra-Levura) whose identity was confirmed by comparing its restriction profile with that obtained previously for an isolate from another commercial batch [44]. This result indicated that the biotherapeutic strain S. cerevisiae var. boulardii was not used in the preparation of the commercial products studied here, thus we decided to include it as positive control in the following section because this strain is positive for the majority of phenotypic traits considered [28]. Only one restriction pattern (A m ) comprised several strains (D2, D3, D4, D7, D8). Another technique, which gives good results in S. cerevisiae clinical strain characterization [44] is PCR amplification of d elements. When this technique was applied, a total of six different amplification patterns were observed ( Figure 1C). Results are summarized in Table 3 by assigning a distinct letter to a pattern when it differs from the others. Of the six strains forming the mtDNA group A m , five were identical while strain D2 grouped differently after PCR amplification. Since different levels of discrimination were obtained, the results of both techniques were combined ( Table 3). As a result, six composite patterns were obtained: five corresponded to individual strains (D2, D5, D6, D14 and D23) and one was a group (A m /B d ) including four strains, three of them isolated from pills (D3, D4 and D7) and one from flakes (D8).
Given that several S. cerevisiae strains were included as controls in the following sections, we considered it of interest to obtain their mtDNA restriction patterns in a way to assure they were clearly distinguishable from the strains under study. As shown in Figure 1B, the strains CECT 10.431, YJM128 and W303 displayed restriction patterns not found in the rest of the strains. The mtDNA restriction pattern of isolate 102 was obtained  previously [44] and is different from that of the strains under study ( [44]: see mtDNA pattern Y m in Figure 1).
3.1. Growth at high temperatures. We analyzed the growth of the commercial yeast isolates at different temperatures (30, 37, 39 and 42uC). Strain D5 (Ultra-Levura), the clinical isolate 102, the laboratory strain W303 and the wine CECT 10.431 strains ( Table 2) were also included as controls in this analysis. Figure 2 shows the growth of the commercial and control strains on YPD plates at different cell concentrations and temperatures. Strains CECT 10.431 and W303 were only able to grow up to 37uC. Meanwhile D4 strain was able to grow at 39uC albeit poorly and only D2, D5 and D14 strains grew at 42uC, as did the clinical isolate 102 whose in vivo virulence was previously shown to be notable since was able to cause death when tested in three murine models of i.v. inoculation [31]. Remarkably, hybrid strains (D6 and D23) only grew at 30uC. The observed growth at 42uC of strain D5 (Ultra-Levura) is in agreement with previously reported data [28].
To analyze growth at 30uC and 37uC in more detail, generation time at these temperatures was determined (Figure 3). Generation time at 37uC in strains D6 and D23 was not determined because these strains grew poorly at this temperature. Strains D2 and D14 had a significantly lower generation time at 30uC and 37uC than avirulent control strain CECT 10.431 and laboratory strain W303, but similar to that of Ultra-Levura (D5) and the clinical isolate 102. Isolate D14 showed the lowest generation times at both temperatures (about half and one third of those corresponding to control strains at 30uC and 37uC, respectively). In contrast to the other strains tested, generation time did not significantly increase at 37uC for strains D14 and 102, indicating that this high temperature is as optimum as 30uC for their growth.
3.2. Extracellular secretion of degradative enzymes. We analyzed the extracellular secretion of phospholipases and proteinases by measuring the precipitation area around each colony resulting from their activity on Egg-Yolk and BSA medium, respectively. The enzymatic activities obtained for each strain expressed as Pz values are shown in Figure 4. All isolates from dietary supplements showed phospholipase activity with levels ranging from low activity (D2, D4, D6 and D23) to moderate activity (D14). Laboratory strain W303 and wine strain CECT 10.431 showed no activity while the D5 strain secreted the enzyme but at low level.
Protease activity could not be assessed for the hybrid isolates (D6 and D23) because, as mentioned above, both isolates were unable to grow at 37uC (optimum temperature to determine this activity). While D4 and D14 showed high activity, as did strain D5 (Ultra-Levura) used as positive control [28], D2 displayed moderate activity levels (Pz = 0.53) showing an intermediate level of activity between the positive (D5) and negative (CECT 10.431) controls. The laboratory strain W303 showed a high Pz value (0.77) corresponding to low activity.
3.3. Pseudohyphal formation. We analyzed pseudohyphal growth and found different degrees of pseudofilamentation in the assayed strains. We considered the absence or presence of this type of growth, morphology and the length of the pseudohyphal chains ( Figure 5). Isolate D2 showed intermediate pseudohyphal growth (Ph +), as did control strain D5 [28], while isolates D4 and D14 exhibited high pseudohyphal development (Ph ++) based on the number and length of filaments. The hybrid isolates (D6 and D23) and the laboratory strain W303 showed an absence of pseudohyphal growth (Ph 2), behaving the same as the negative control (CECT 10.431).
3.4. Invasive growth. Some S. cerevisiae isolates are capable of penetrating agar, producing filaments able to resist vigorous washing of the surface of solid growth medium. This feature is known as invasive growth [56].
Three different behaviors regarding invasive growth were observed in the assayed strains, based on the residual growth left after washing the cultures grown on YPD plates for five days at 30uC ( Figure 6). Isolates D4 and D23, together with the laboratory strain W303 and the control strain D5, were fully eliminated after washing (negative invasive growth: -). D2 and D14 remained partially anchored to the agar, forming a thin layer of yeast (intermediate invasive growth: 6); this type of growth was also shown by the control strain CECT 10.431. Only isolate D6 showed positive invasive growth (+), with practically all the culture being attached to the agar surface after washing.
3.5. Colony phenotype switching. The switching phenomenon, described in some fungi like C. albicans, C. tropicalis or C. neoformans, is a mechanism of phenotypic plasticity manifested by spontaneous reversible changes in colony morphology occurring at a frequency of about 10 23 -10 24 . We studied this phenomenon in four of the strains isolated from dietary supplements (D2, D4, D14, and D23) by analyzing the phenotypic changes in a number of colonies between 6000 and 8000. All four strains showed a main phenotype and several secondary phenotypes; the latter with a frequency of appearance between 3.4610 22 and 1.91610 24 (Table 4). Specifically nine colony phenotypes were observed in D2 strain, six in D4 strain and eight in strains D14 and D23. An example of the colony phenotypes obtained for strain D2 and D14 is shown in Figure S1. Stability and reversion of some of these secondary phenotypes were also studied. There were reversible and stable secondary phenotypes in the four strains analyzed (data not shown). The stable secondary phenotypes were able to revert to the main phenotype at low frequency, except two secondary phenotypes of strains D4 and D14. These results indicate that these strains showed colony switching following a similar pattern.
3.6. Activation of MAPK signaling pathways. To study the basal level of activation of the pathways mediated by Slt2 and Kss1 in dietary and probiotic Saccharomyces strains, the dually phosphorylated forms of these proteins were detected by immunoblotting with polyclonal anti-phospho-p44/p42 MAPK (Thr-202/Tyr-204) antibody that specifically recognizes activated MAPKs [51]. Anti-Slt2 and anti-Kss1 antibodies were used to detect the total amount of these proteins. Western blotting was performed with extracts of the different strains grown at 24uC, in the absence of external stimulating conditions of these MAPKs signaling pathways. Extracts of the laboratory strain BY4741 grown at 24uC and 39uC were used as controls respectively for basal phosphorylation of these MAPKs and activation of Slt2 upon stimulation of the CWI pathway, as in a previous work [30]. Differences in phosphorylation of Slt2 and Kss1 were observed among the different strains tested (Figure 7). Greater phosphorylation of Slt2 was observed in strains D5, D6 and D14 as compared to control laboratory strain under the same conditions. An increased level of phospho-Kss1 was only detected in strain D5 (Ultra-Levura), in agreement with a previous study [30].
3.7. Adherence to polystyrene plastic. In Candida spp. adherence to host tissues and plastic surfaces has been associated with pathogenesis, because it is the first step in colonization and development of infection [25,26]. For this reason, we analyzed the adherence of the Saccharomyces strains under study to different plastics. To test adherence to polystyrene, two materials were used: petri plates and 96-well microtiter plates. Laboratory strains BY4741 and W303 were used as negative controls and strain SC5314 of C. albicans was chosen as positive control for adherence. We also tested the S. cerevisiae clinical isolate 102. As shown in Figure 8, strain D5 showed very low adherence to Petri plates (even lower than control strain W303), whereas strains D2, D4 and D6 showed similar adhesion to those displayed by the clinical strain 102 and the C. albicans strain SC5314. Strains D14 and D23 showed lower adhesion rates, similar to that of laboratory strain W303 and greater than that of laboratory strain BY4741 (Figure 8). In the adherence assay on 96-well microtiter plates strain D6 showed the greatest adhesion among the dietary and probiotic strains while the other strains gave values equal to or lower than the laboratory control strains. Remarkably, whereas strain 102 displayed high adherence in this assay, the C. albicans strain did not show greater adherence than laboratory strains of S. cerevisiae.
In summary, high adherence to polystyrene was observed for strains D6 and 102 with both methods, while strain D5 showed the lowest adhesion rates.

3.8.
Adherence to polyurethane intravenous catheters. Intravenous catheters constitute a route of entryof microorganisms into the bloodstream. Thus, we decided to test the adherence of the D14 strain to intravenous catheters because its virulence in murine model has been described [37] and the 102 clinical strain due to its observed high adhesion to polystyrene. We also used the SC5314 strain of C. albicans as a positive control of adhesion. These experiments were performed on polyurethane catheters used for critically ill patients (long-term catheterized). Glucosamine solution was used because this medium mimics the As shown in Figure 8C, the strain D14 showed low adherence in both experiments. In contrast, the clinical S. cerevisiae isolate 102 and the C. albicans SC5314 strain displayed similar high adherence  to polyurethane in glucosamine solution, which is consistent with their ability to adhere to polystyrene, although the adherence of the 102 strain was lower than that of C. albicans when RPMI is used. After 24 h in such medium, C. albicans was the only strain able to form a biofilm adhered to catheters. For this reason it was not possible to quantify the number of adhered cells, although biofilm formation implies high adhesion ability.

In vivo Virulence by Intravenous Inoculation in Mice
The results of the analysis of virulence-associated phenotypes (summarized in Table 5) show that isolates D2, D4 and D14 displayed positivity for 63.6% to 72.7% of the phenotypic traits studied (Table 5). Specifically, the three isolates were able to grow at 37uC and 39uC, showed pseudohyphal growth and secreted proteases and phospholipases. The only difference was the ability to grow at 42uC which was negative for D4 and activation of MAPK Slt2 that was positive only for D14. The association perceived between D2, D4 and D14 based on their in vitro behavior was shown positive and statistically significant in terms of the Spearman's correlation coefficient (p = 0.01 between D2 and D4; p = 0.030 between D2 and D14) and this result is consistent with the analysis performed using the Jaccard Proximities test that displayed the maximum similarity values (between 0.700 and 0.889) for these three strains. Furthermore, D2, D4 and D14 were clearly differentiated from the two hybrid strains (D6 and D23) (p.0.05, Spearman's correlation coefficient; similarity values between 0.182-0.250, Jaccard Proximities test). Taking into account that these results indicated that the group of isolates named D2, D4 and D14 could be potentially virulent, we decided to study their virulence in vivo. Moreover, we considered of interest to determine if there was association between absence of virulenceassociated phenotypes and pathogenicity in vivo. To do so, we also tested D23 as it was the isolate with the lowest percentage of positive traits (20%) compared to D6 (44.4%).
We evaluated the potential ability of D2, D4, D14 and D23 to develop systemic infection after their intravenous inoculation (i.v.) in immunocompetent BALB/c mice and we compared the results with those of the laboratory strain W303, the wine strain CECT 10.431 and the clinical strain YJM128. The strains CECT 10.431 and YJM128 were chosen as non-pathogenic and pathogenic controls because they displayed low and high capacity to lodge and proliferate in representative organs respectively, after intravenous inoculation of mice [16,31,37].
Pathogenicity was evaluated according to death rates and by determining colonization ability (the presence of yeast cells in brains and kidneys as demonstrated by culture). Three days after intravenous inoculation, mice inoculated with strains D2, D4, D14, and the virulent control YJM128 strain showed signs of infection, in particular, weight loss, hair frizz and low mobility, which could indicate a febrile process. After the first seven days, most of affected mice recovered, except an average of 27.5% of those inoculated with isolate D14, which died between the second and fourth days post-infection ( Figure 9A). The difference of mortality observed between mice inoculated with D14 and the rest of isolates was statistically significant (p,0.05, log rank and Wilcoxon tests). Therefore, D14 was the only commercial isolate causing lethality among the assayed strains. As observed in Figure 9B, the number of dead mice due to i.v. inoculation of D14 varied depending on the assay. The differences in mortality rates found between the four trials were equivalent (p.0.05, log rank and Wilcoxon tests).
The differences between isolates became evident at day 15 postinfection. Therefore, burdens at this time provide information about the clearance profile of each strain and hence their pathogenic potential. As shown in Figure 10, the burden values in brains at this time were significantly lower than those obtained on day 7 in this organ for all strains (p,0.05, one-way ANOVA followed by Tahmahe's post-test) except for D14, indicating that D14 is the one with the greatest resistance to clearance. D14 is also remarkable because it shows significantly higher burden levels than the other isolates, including the positive control (YJM128) (p,0.05, one-way ANOVA followed by Tahmahe's post-test). Moreover, burden values shown by D2 and D4 in brains were equivalent to those shown by the positive control (YJM128) (p. 0.05, one-way ANOVA followed by Tahmahe's post-test), and clearly higher than those of the negative control (CECT 10.431) (p,0.05, one-way ANOVA followed by Tahmahe's post-test). However, at day 15 post-infection, kidney clearance was particularly evident in D2 and D4, as both isolates showed undetectable burdens as did the negative control. Again, D14 stood out since it was the only dietetic isolate able to persist in this organ and at levels equivalent to those shown at day 7 (p.0.05, one-way ANOVA followed by Tahmahe's post-test). This pattern of behavior is comparable to the strain used as positive control (YJM128). By day 30 post-infection, the majority of isolates had been completely eliminated in both organs except strains D14 and YJM128, which were detected at statistically insignificant levels Table 5. Virulence-associated phenotypes observed in strains isolated from commercial products.    Phenotype traits shown in Figures (2, 4, 5, 6, 7, 8) and Table 4 were simplified as follows to calculate the Spearman's Rho correlation coefficients between strains and apply the Jaccard proximity test (excluding joint absences for considerations): growth at 37, 39 and 42uC,  compared to the negative control (p.0.05, one-way ANOVA followed by Tahmahe's post-test).
In summary, according to the degrees of virulence proposed by Byron et al. [15] based on burden levels and ability to cause death, D14 could be considered a virulent isolate; D2 and D4 isolates, despite showing high burden levels in brains at 7 and 15 days postinfection, were unable to cause death and, therefore, can be considered of low virulence; D23 together with laboratory strain W303 can be considered avirulent, given their inability to colonize BALB/c mice after intravenous inoculation.

In vivo Virulence by Oral Inoculation in Mice
The in vivo virulence study by i.v. inoculation described in the previous section has shown the potential pathogenicity of some Saccharomyces strains isolated from dietary products. It should be noted that due to the oral entry of yeast cells contained in such preparations, only those strains capable of passing through the gastrointestinal barrier and spreading are potentially dangerous. According to death rates and burden data obtained by i.v. inoculation, D14 is the isolate posing the greatest risk. Therefore, its infectivity was assessed in an experimental murine model of gastrointestinal infection with immunosupression and disruption of mucosal integrity, and compared with the avirulent D23 isolate.
The presence of yeasts in the intestinal tract was evaluated by determining fecal load from the beginning of administration (day 0) to the end of the trial (day 7) (Figure 11). One day after oral challenge, stool cultures from all mice tested positive for both strains. Yeast levels remained at values between 6.5 and 7.3 log 10 CFU/gram of feces for D14 and between 4.9 and 6.8 log 10 CFU/ gram for D23, during the entire experiment even after stopping the administration of the corresponding strain in drinking water (from day 3). According to these data, the treatment with an antibiotic regimen kept D14 and D23 in the gastrointestinal tract at levels in the same order as those achieved by prior researchers for other yeast species [57,58] and, therefore, they were sufficient to undertake the study of intestinal infection.
Antibiotic administration was maintained throughout the experiment to minimize complications of bacterial infections occurring as a result of dexamethasone treatment. Although the use of antibiotics did not completely eliminate the bacterial flora (data not shown), residual bacteria did not have a significant pathogenic effect, as indicated by the fact that control mice that were not inoculated with yeast cells neither showed signs of infection nor died.
One day after gastric inoculation (day 1), only mice inoculated with the D14 strain presented ruffled hair, symptom of an infection process, but retained good mobility and experienced no weight loss. These characteristics were maintained for the remainder of the experiment, being more pronounced in the last days. First, we evaluated the translocation of D14 and D23 from the intestinal tract by assessing burden levels in Peyer's patches; both isolates crossed the intestinal barrier in the experimental conditions although translocation was more consistent in the case of the virulent isolate (D14) as it was detected in Peyer's patches from 100% of mice tested, while the avirulent isolate (D23) was detected in 75% of them (Table 6). Remarkably, burden levels of isolate D14 in this organ were higher than those of isolate D23 and these differences were statistically significant (p,0.05, one-way ANOVA analysis) ( Figure 12). In regard to dissemination, differences between both isolates were evident. D14 spread to MLNs in 33% of mice and to distant organs (livers, kidneys and brains) in 75%. Conversely, D23 was not found in MLNs or brains and was occasionally found in livers and kidneys (in 25% of total mice and only at day 5 post-inoculation).

Identity of the Yeast Cells Recovered from Infected Mice
In order to confirm that the colonies recovered from the infected mice really belonged to the yeast strain inoculated, mtDNA restriction analysis was performed both in the intravenous and oral models of infection.
The mtDNA pattern of each recovered colony form each organ was compared with that displayed by the strain inoculated in each mouse, and 100% of identity was found in all cases (data not shown).

Discussion
In recent years, the perception that S. cerevisiae as harmless has changed due to an increase in the number of infections caused by this yeast [6,59,60]. There is no doubt about the opportunistic nature of these infections; however, there are studies indicating that not all strains are capable of developing infection, even though favorable conditions are present [15,16,31]. Regarding S. cerevisiae strains contained in dietary supplements, it is unknown whether they have intrinsic properties that could contribute to virulence in humans. Given that these commercial preparations constitute the main route by which living S. cerevisiae cells gain entry into the human body, the study of their in vivo virulence is of great importance.
When the systemic virulence of dietetic Saccharomyces isolates was assessed by intravenous inoculation, we found that among the isolates tested D2, D4 and D14 were able to lodge and persist in the analyzed organs (brains and kidneys) while isolate D23 was undetectable. The ability to lodge has previously been described for a large majority of S. cerevisiae clinical isolates [15,16,31,37] and for some non-clinical isolates [16,17,31,37]. Burden values obtained for our isolates were higher or similar to those observed for the positive control (YJM128) although this strain was previously described as highly virulent in an immunocompetent murine system (CD-1) [16]. This is probably due to the fact that a yeast strain can exhibit different degrees of virulence depending on the kind of mouse used to mimic the systemic infection [15]. Indeed, YJM128 was more virulent in DBA/2N mice than in CD-1 mice [15,16] and, in the light of our results, it appears to be less virulent in BALB/c mice. Moreover, a loss of virulence with the passage of time cannot be ruled out, as described by Clemons et al. [16] for another isolate, and it must be taken into account that YJM128 was first described in 1994. Despite this observation, we have referred to YJM128 as a positive control of virulence when intravenous inoculation was used as the infection route.
One of the most remarkable results of this work was the ability of a strain isolated from a commercial dietary supplement (the D14 strain) to cause mortality in BALB/c mice. This result confirms previous data reported by our group indicating a remarkable virulence of this strain using other mouse strains (DBA/2 and C3H/HeN) [37]. The null mortality observed in one of the infection assays with D14 could be a result of increased resistance to infection associated with a mature immune system in this particular batch of mice. In a previous study, we have shown the influence of immune system maturity on the virulence of S. cerevisiae strains when BALB/c mice were used [31].
By contrast, isolates D2, D4 and D14 produced clear signs of disease in mice, like the YJM128 strain. The same signs of disease were described in the course of C. albicans-related infections, a typically pathogenic yeast strain [61]. Therefore, this information would indicate that these S. cerevisiae isolates not only can survive in mouse blood and then spread to different organs, but they are also able to trigger a systemic infection. This finding is important because colonization is not always connected with symptoms [31] and highlights the particularity of these isolates.
Once host-yeast interaction takes place, the expression of virulence factors and the ability to proliferate and survive in the host or to cause damage could differ from in vitro conditions. That is why it is questionable whether there is an association between the phenotypic traits expressed by the isolates and their virulence in vivo. In our study, we have found data suggestive of a correlation between some of the virulence-related traits in vitro and the results obtained in vivo in a murine model of i.v. inoculation, since the strains thought potentially pathogenic based on phenotypic traits (D2, D4 and D14) were able to lodge and persist for some time in the analyzed organs and developed symptoms of infection. Moreover, the isolate D23, which only showed a low phospholipase activity among the virulence-related traits assayed, was found to be completely avirulent in vivo.
We conducted a statistical analysis including all the strains whose in vivo virulence was tested (dietetic strains and control strains) to evaluate this hypothesis (Table 7) and we found that the ability to grow at 39uC and pseudohyphal growth are positively and very strongly associated with in vivo virulence in BALB/c mice (Spearman's correlation coefficient = 1, p,0.01; similarity values = 1, Jaccard Proximities test). These traits have previously been studied in clinical and non-clinical S. cerevisiae strains [28] and have been associated with the in vivo virulence of S. cerevisiae [11,31]. Some considerations can be made in relation to growth at .37uC. We noticed that the association between the ability to grow at 42uC and the in vivo virulence was in the boundaries of significance (p = 0.052, Spearman's correlation coefficient, Table 7) suggesting that both traits are close. This result resembles that shown by McCusker et al. [11] although the latter author reported a strong association between the ability to grow at 42uC and the ability to colonize immunocompetent CD-1 mice. This discrepancy is probably due to the type of strains they tested. They analyzed both clinical and non-clinical strains and show that the ability to grow at 42uC was mainly related to the clinical origin. Indeed, it seems logical that microorganisms with difficulties to grow at human body temperature or those reached during febrile states, are not able to develop systemic infection, as observed for strains D6 and D23. The inability of these strains to grow at temperatures $37uC is probably related to their hybrid nature and habitat since S. cerevisiae-S. bayanus and S. cerevisiae-S. kudriavzevii hybrids are primarily associated with alcoholic fermentations carried out at low temperatures [62]. Therefore, it seems unlikely that the hybrid nature facilitates colonization, invasion and dissemination during a systemic infection. The opposite occurs at the industrial level, since hybrids are better adapted than their parental strains to resisting stress during fermentation processes [41]. According to these comments hybrids could be safer than S. cerevisiae strains for the preparation of dietary supplements. However, it would be necessary to extend this study to a larger number of hybrid isolates to confirm this hypothesis.
Protease and phospholipase production, invasive growth and colony switching, have been associated with clinical origin but they are less obvious virulence-related attributes [12,16,28,63]. This last statement is consistent with our statistical analysis of association (Table 7). We did not observe any significant association in the case of MAPK activation and plastic adherence either, even though both traits are important pathogenic factors in C. albicans [25,26,27]. It is worth commenting that the increased phosphorylation of the MAPK Slt2 would reflect an activation of the CWI pathway in the virulent D14 strain that could be due to an altered cell wall. The alteration of this external structure in yeast cells might affect immune system recognition, as proposed for a S. cerevisiae hypervirulent mutant defective in the SSD1 gene [64]. These authors suggest that the altered cell surface of this mutant might lead both to misrecognition by the innate immune system and induce shock in the mice by overstimulation of the proinflammatory response. Furthermore, the surface-adhering ability observed in D2 and D4, with intermediate virulence level after intravenous inoculation, could facilitate their entry to patients' bloodstream and internal organs and would, therefore, make them more infectious in theory than other strains lacking this ability. Strain D5 (Ultra-Levura) showed the lowest adherence to any of the plastics tested among the analyzed strains This was unexpected given given its virulence in experimental models [17,31] and its association with human infection via contamination of central venous catheter by the colonized hands of health-care workers [6,59]. In constrast, the S. cerevisiae clinical isolate 102 showed significant plastic adhesion ability, comparable to C.albicans for polyurethane catheters and ever higher to that displayed by this pathogenic species to polystyrene.
As we have stressed, there are Saccharomyces isolates from dietary products able to colonize and disseminate and even cause death in animal models when inoculated directly into the blood. However, since these yeast-cell-containing products are ingested orally, infection models derived from translocation across the gastrointestinal mucosa are required to determine which strains may pose a real potential safety risk to humans. To date, the few studies into the occurrence of S. cerevisiae intestinal translocation are limited to investigating its probiotic properties; in particular the protective capacity against infection by pathogenic bacteria and rotavirus [65][66][67][68][69][70]. There is only one study investigating the virulence of S. cerevisiae when administered as a probiotic [71]. Given that isolate D14 was highlighted for its virulent potential in an i.v. inoculation model, we considered it important to assess the negative impact of its consumption. In clinical cases of fungemia derived from the intake of live S. cerevisiae cells, in particular S. boulardii, it has been postulated that translocation from the gut to the bloodstream occurs in debilitated patients with impaired gastrointestinal mucosa, treatment with broad-spectrum antibiotics and/or severe immunosuppression [34,35,72,73]. The same behavior can be expected in the case of S. cerevisiae isolates from dietary supplements, justifying the use of animal systems that mimic the abovementioned conditions. Moreover by using an in vitro intestinal epithelial barrier model, we have previously shown that, in contrast to C. glabrata and C. albicans, S. cerevisiae was not able to cross the intestinal barrier unless its integrity was compromised [74]. Thus, a murine model has been established using mice pretreated with a combination of three antibiotics (streptomycin, gentamicin and bacitracin) and the administration of high-doses of dexamethasone, which has enabled intestinal infectivity of isolate D14 to be studied and compared with the isolate considered avirulent (D23).
The burden levels detected in Peyer's patches, the first organ where the pathogenic agents crossing the gut wall can be phagocyted, show the occurrence of intestinal translocation in sub-mucosa in both isolates, although it was more consistent for D14. Translocation has already been described to occur in probiotic strains of S. cerevisiae during studies of its immunomodulating action or competition with harmful bacteria and rotavirus [65][66][67][68][69][70]. In some of these studies, the presence of yeast was detected in mesenteric nodes of immunocompromised mice [68], young mice (21 days) [69,70] or infant mice [65,66]. However, studies investigating the spread of yeast cells to distant organs did not report the presence of yeast cells in such organs [70,71] or their presence was inconsistent and occasional [65,68]. These negative results were probably due to the use of inappropriate murine model that did not combine immunosuppression and intestinal trauma. In our study, a four-day application of dexamethasone, a drug that leads to immunosuppression, besides damaging the gastrointestinal mucosa, allowed D14 to spread to mesenteric nodes and distant organs. The fact that some mice showed no burdens in mesenteric nodes but in distant organs could indicate the occurrence of direct vascular invasion with subsequent yeast transport through the portal system to the liver and, hence, to other distant organs besides spreading via the Table 7. Statistical analysis of association between virulence-associated phenotypic traits and in vivo virulence.
In vivo virulence b Traits ordered by proximity for better comprehension. Data used in the case of strains isolated from dietary supplements are shown in Table 5. Phenotypic traits of control strains, simplified following the codes in Table 5, are: YJM128 (+, +, +, +, -, +, +, n.d., n.d., n.d., n.d.), W303 (+, -, -, +, -, -, -, n.d., n.d., n.d., -), CECT 10.431 (+, -, -, -, -, -, -, n.d., n.d., n.d., n.d.). Data of YJM128 have been obtained previously [28]. ''+'' and ''-'' were coded numerically as ''1'' and ''0'' respectively for statistical analysis. b In vivo virulence determined by intravenous inoculation of strains D2, D4, D14, D23, YJM128, W303 and CECT 10.431 in BALB/c mice. Virulence was simplified as follows: high burdens levels in brain (3.18 to 5.31 mean log 10 CFU) at 7 and 15 days post-infection were considered + and lower burdens levels and the inability to colonize was considered -. c A proximity value of ''1'' indicates the maximum association and value ''0'' indicates the opposite. Jaccard proximities were calculated excluding joint absences from consideration. d Values of ''1'' indicates perfect correlation and ''00 indicates no relationship. n.c. Spearman's correlation cannot be calculate because this trait shows constant values (see Table 5). doi:10.1371/journal.pone.0098094.t007 lymphatic system. The results obtained for isolate D14 contrast with the behavior of D23 and agree with the virulence degree assigned to both isolates after the i. v. infection study. Therefore, in the presence of conditions promoting intestinal translocation, the spread to internal organs appears to be highly strain-dependent and only seems to occur when it possesses innate virulence attributes; however, we cannot generalize since only two strains have been tested. This is in agreement with a study using piglets orally inoculated with C. albicans, in which a non-virulent strain showed poorer dissemination than a virulent one [75].
To our knowledge, this study is the first to demonstrate the occurrence of virulence-related traits in Saccharomyces strains ingested as living cells in the form of commercial dietary products. Moreover, we show that some of these isolates can be pathogenic to animal models when inoculated intravenously and at least one of them, when the inoculation route is oral and the mice have an impaired gastrointestinal mucosa. Taking into account that dietary products could be a natural way for live yeast cells to gain entry into the human body, the strains used in their production should be avirulent. Therefore, some recommendations should be made to yeast production and distribution companies and dietary supplement manufacturers. On the one hand, prior screening of virulence-related attributes, such as those tested here (particularly growth at 39uC and pseudopyphal growth), appears to be necessary when selecting new strains. On the other hand, the health foods currently available on the market should also be checked to ensure their safety. All these precautions are especially important for immunocompromised individuals. Furthermore, the availability of virulent and avirulent isolates is very useful for future works aimed to explore the mechanisms involved in S. cerevisiae virulence as well as in other emerging food-borne yeast pathogens. Figure S1 Example of colony phenotype switching. Representative colonies of D2 (A) and D14 strain (B) after five days at 30uC on YPD with phloxine B. (TIF)