INTRODUCTION

Top Abstract Introduction Materials and Methods Results Discussion References

The hemiascomycetous yeast Saccharomyces cerevisiae has been used for centuries for the production of fermented food such as cider, wine, beer, and bread. More recently, it has also been considered as a nutritional supplement specially used by sportsmen. Saccharomyces boulardii, now considered a strain of S. cerevisiae (17), is prescribed as a biotherapeutic agent for the treatment or the prevention of diarrhea, notably that related to Clostridium difficile or associated with enteral and parenteral nutrition (20).

It is not known if S. cerevisiae is a true commensal of the human intestinal flora or only a transient agent of gastrointestinal colonization, possibly related to its ingestion with food. Yet over the past several years S. cerevisiae has been identified as an occasional human pathogen. Virulence traits have been documented both in vitro (pseudohypha formation, growth at 42°C) and in animal models (18). In women, S. cerevisiae is mainly responsible for vaginitis, clinically comparable to recurrent candidal vaginitis, which can be effectively treated by the extensive use of azole derivatives, against which S. cerevisiae is usually resistant (27, 30). S. cerevisiae may also be responsible for severe disease in immunocompromised patients (1, 7) and is now considered an emerging opportunist pathogen (12). Over the past decade, a growing number of reports on systemic and disseminated infections due to Saccharomyces have been published, and a recent epidemiological study showed that S. cerevisiae was responsible for 3.6% of all fungemias (24). It has been recently demonstrated that some of these infections are related to S. boulardii therapy (13, 24), but the epidemiology of human Saccharomyces infections remains poorly understood.

Thus, there is a need for molecular markers able to distinguish strains. Despite the availability of several molecular methods, data on S. cerevisiae typing are still limited. Moreover, some of these methods, such as karyotyping or mitochondrial DNA polymorphism, are incompletely evaluated because of the low number of strains tested or because of the lack of studies on stability and reproducibility (28). In addition, the discriminatory power of some methods appears insufficient when these methods are tested alone. In a comparative study of several techniques including randomly amplified polymorphic DNA, PCR fingerprinting, and enzymatic restriction of amplified DNA, Baleiras Couto et al. concluded that there was no single PCR-mediated typing technique able to discriminate at the strain level (2). More-laborious pulsed-field gel electrophoresis of NotI-digested DNA appeared to be a valuable typing method since it could distinguish 62 distinct patterns from 76 clinical isolates (30). Restriction fragment length polymorphism (RFLP) generated by EcoRI digestion of total DNA allowed the differentiation of 41 subtypes among 60 isolates (49 clinical and 11 nonclinical) (6). Moreover, the patterns could be divided into two genetic groups that were distinguishable by the virulence of the strains when tested in a murine model (5).

Microsatellites are short (usually less than 10-bp) sequence repeats which have been shown to exhibit a substantial level of polymorphism (~10² to 10⁵) in a number of eukaryotic genomes (26). In humans, they have been used in paternity tests, forensic medicine, and population structure studies. They have been successfully applied for typing fungi such as Candida albicans (16) and Aspergillus fumigatus (3). The complete sequence of the S. cerevisiae genome allows the identification of these regions and thus their use for the development of this novel molecular tool for typing. Here, we report the use of microsatellite polymorphism as a new tool for the identification of S. cerevisiae strains. It is shown that the patterns are stable and that the method is discriminant, making it a powerful tool for epidemiological purposes.

	MATERIALS AND METHODS

Top Abstract Introduction Materials and Methods Results Discussion References

Strains. Ninety-one strains were tested including 14 laboratory strains among which was the S288c strain, which has been completely sequenced (11), 28 strains with industrial relevance collected in eight countries, i.e., France, Spain, Russia, China, Taiwan, Vietnam, Japan, and the Czech Republic, seven reference strains of unknown origin, and 41 isolates collected from 36 patients (Table 1). These patients were either outpatients or hospitalized in eight different institutions from three European countries (France, Spain, and The Netherlands). Isolates were collected from routine cultures of vaginal discharges, stools, sputum, bronchoalveolar lavages, and blood cultures. Nineteen patients (21 isolates) were receiving S. boulardii therapy when their isolates were collected. Additionally, three isolates of S. boulardii (one reference strain and two isolates collected from commercial preparations) were tested. Strains from closely related species, i.e., Saccharomyces pastorianus (CLIB 176, CLIB 180, CLIB 277, CLIB 1260, and CLIB 1486), Saccharomyces paradoxus (CLIB 97, CLIB 228, CLIB 2980, and CLIB 7400), and Saccharomyces bayanus (CLIB 251 and CLIB 106) were also tested. Strains were stored at 80°C in 10% glycerol until tested.

Microsatellite polymorphism. (i) DNA extraction. DNA was extracted according to the method previously described (15). Briefly, yeast cells were grown overnight at 30°C in 3 ml of YPGlu medium (1% [wt/vol] yeast extract [Difco Laboratories, Detroit, Mich.], 1% [wt/vol] Bacto peptone [Difco Laboratories], 2% [wt/vol] glucose). Yeast cells were pelleted and resuspended in 200 µl of lysis buffer (50 mM Tris-HCl [pH 8], 25 mM EDTA, 1% [vol/vol] -mercaptoethanol, 100,000 U of Zymolyase [ICN Pharmaceuticals, Costa Mesa, Calif.]). After 1 h at 37°C, 200 µl of a solution containing 200 mM diethanolamine, pH 9 (Sigma), 80 mM EDTA, pH 9, and 1% (wt/vol) sodium dodecyl sulfate was added and the mixture was incubated at 65°C for 30 min followed by 5 min on ice. A volume of 100 µl of potassium acetate, 5 M, was added at 4°C for 45 min. After centrifugation for 5 min at 10,000 × g, the supernatant was precipitated by adding ammonium acetate at a final concentration of 2.5 M plus 3 volumes of 100% ethanol. DNA was pelleted by centrifugation for 10 min at 10,000 × g, rinsed with 70% ethanol, dried, and resuspended in 50 µl of Tris-EDTA buffer.

(ii) Microsatellite amplification. Seven microsatellites (five trinucleotide and two dinucleotide simple repeats) were selected (Table 2). Oligonucleotide primers were designed from the sequences of the flanking regions to obtain PCR products ranging in size between 110 and 170 bp. Amplification was performed with a mixture containing [-³²P]dATP (500 µCi), 500 mM KCl, 100 mM Tris-HCl (pH 8.3), 15 mM MgCl₂, 0.1% gelatin, deoxyribonucleotide triphosphate (0.2 mM), forward and reverse primers (20 pmol each), Taq polymerase (1.5 U; Eurogentec, Seraing, Belgium), and 0.5 to 1 µg of DNA. Amplification was achieved through 30 cycles (94°C for 15 s, 55°C for 1 min, 72°C for 1 min 30 s) and a final extension step at 72°C for 10 min.

(iii) Length polymorphism analysis. Radiolabeled PCR fragments were electrophoresed in 7.5 M urea-12.5% (1:30) polyacrylamide vertical gels at 90 mA for 3 h. Length polymorphisms were evaluated by comparison with the migration of the S288c reference strain tested in triplicate in each run.

Evaluation of the method. To test genotype stability, DNA from six independent subclones of S288c strain were extracted and tested simultaneously for the seven loci. In addition, an S288c strain was grown in a fermentor for 35 weeks (about 690 generations). Twice a week, aliquots were collected and DNA was extracted and tested by PCR for three loci (loci 1 to 3). Genotypes of strains 25 to 32 were determined and compared to results of the same analysis performed 3 years ago. The reproducibility was assessed by testing the S288c strain 12 times in four separate experiments for each microsatellite. Also, S. boulardii strain UL G84F88I90 was included in each of these experiments.

Phylogenetic analyses. We considered that the number of repeats at each locus was equiprobable for each isolate. Thus, all alleles were considered as phylogenetic markers with two character states, i.e., present (1) and absent (0). The matrix was analyzed using the maximum-parsimony approach (unweighted heuristic search; 100 random stepwise addition sequences) with PAUP 4 software (29). A strict consensus tree was drawn using the midpoint rooting option. Statistical robustness of nodes was evaluated using bootstrap resamplings (100 iterations). Because of computation time the "fast stepwise addition" option was used for each replicate. The consensus tree is given, with branch lengths obtained using the Minf optimization.

Statistical analysis. To test the hypothesis of a nonrandom distribution of the allelic sizes, we calculated for each locus the chi-square value by comparison of the observed distributions to theoretically normal distributions characterized by the same geometric mean and standard deviation. A P value below 0.05 was considered significant.

	RESULTS

Top Abstract Introduction Materials and Methods Results Discussion References

All the S. cerevisiae strains, including S. boulardii, gave 100% positive PCR results with all microsatellites tested. In contrast, typeability, defined as the proportion of strains that are assigned to a type, was 0 for S. bayanus and ranged between 0.25 and 0.5 for S. paradoxus and between 0 and 1 for S. pastorianus, depending on the locus tested. Each locus appeared polymorphic with 12, 13, 6, 11, 8, 6, and 13 alleles for microsatellites 1 to 7, respectively (Table 3). For each locus considered and each strain, the size of the microsatellite on gel was estimated. Figure 1 gives an example of the stability of the patterns and the discriminatory power of the method. In some of the strains analyzed, we found two bands at a given locus. We concluded that these strains were diploid and heterozygous for the allele considered (Table 3). In contrast, when an isolate exhibited only one band for all loci, no conclusion on the ploidy of the isolate could be drawn. For different loci, allele sizes were distributed somewhat differently (data not shown). In some cases, one allele was predominant and smaller or larger alleles were found in the analyzed population, demonstrating a progressive continuum for increasing or decreasing numbers of repeats (microsatellites 2, 5, 6, and 7). In other cases, no predominant allele was observed but the sizes were distributed among six or more categories (microsatellites 1, 3, and 4). However, none of the loci studied demonstrated what could be considered a normal (Gaussian) distribution (P < 0.05) (data not shown).

View larger version (48K): [in this window] [in a new window]   FIG. 1.   Denaturing polyacrylamide gel electrophoresis of products of PCR amplification of the YKL172w microsatellite (microsatellite 1). S. cerevisiae strains are described in Table 1. Note that CLIB 1260 (lane 3) corresponds to an S. pastorianus strain. *, clinical isolates that were collected from patients taking S. boulardii therapy.

A total of 69 different alleles were observed, generating 52 distinct patterns. Pattern identity was observed for French strains used in the agroalimentary industry: strains 54 and 55; strains 44, 45, and 47; strains 46, 48, and 50; and strains 49, 52, and 53. Similarly, among laboratory strains, strains 30 and 31 and 10 and 11 were indistinguishable. The last two were indistinguishable from an apparently non-epidemiologically related strain (strain 24).

The patterns of the 21 isolates collected from the patients receiving S. boulardii therapy were indistinguishable. These patterns were also indistinguishable from those of the two isolates of S. boulardii collected from commercial preparations and the reference strain from the manufacturer. S. boulardii-related isolates exhibited an allele for locus 4 that was not shared with any other strain. Among the 19 isolates collected from 17 patients not taking S. boulardii therapy, we found 14 unique patterns. Four French clinical isolates gave a pattern indistinguishable from that of a French bakery strain (strain 2). There were two pairs of clinical isolates that could not be differentiated. In both cases, they were isolated from the same patient (isolates 66 and 67 and 58 and 59). None of the patients not treated with S. boulardii had an isolate with a pattern characteristic of S. boulardii.

Performance of the method. Genomic stability of assayed microsatellites was confirmed by analysis of six independent subclones of the S288c strain. In addition, no change of genotype was observed over the 690 generations collected during 35 weeks of growth in a fermentor (see Materials and Methods). Moreover, stability was also confirmed by comparison between experiments performed 3 years ago and the present study. Independent PCRs run using the same DNA (strain S288c and UL G84F88I90) and the same primers gave the same pattern, demonstrating the total reproducibility of the technique. The discriminatory power calculated for a sample of 58 strains considered unrelated was 99.03%.

Phylogenetic analysis. The phylogenetic analysis provided 1,479 equiparsimonious trees of 208 steps, with a consistency index of 0.332 and a retention index of 0.7 (Fig. 2). Four clades exhibited bootstrap proportions over 70%: the clades comprising strains 2, 5, 7, 8, and 56 (95%); strains 58 and 59 (85%); strains 66 and 67 (99%); and strains 54 and 55 (95%). Also, it was noted that 12 of the 14 laboratory strains form a clade. Similarly, 8 (strains 34, 36, 37, 38, 39, 40, 41, and 42) out of 10 Asian isolates are members of the same clade. Finally, 7 (strains 43 to 48 and 50) out of 11 French wine strains showed a strong identity. However, it was not possible to distinguish a clade comprising clinical isolates. Also, there was no difference between strains isolated in true pathogenic conditions (isolated from blood and cases of vaginitis) and those isolated as probably commensal organisms (feces).

View larger version (24K): [in this window] [in a new window]   FIG. 2.   Strict consensus of the 1,479 most-parsimonious phylogenetic trees. The phylogram is constructed using the midpoint rooting method. Tree length, 208; consistency index, 0.332; retention index, 0.7. Strain numbering refers to Table 1. Clinical isolates from patients not taking S. boulardii are circled. * (boxed), S. boulardii reference strain (1), isolates from S. boulardii commercial preparations (90 and 91), and isolates from patients treated with S. boulardii (69 to 89).

	DISCUSSION

Top Abstract Introduction Materials and Methods Results Discussion References

The increasing importance of fungi among nosocomial infections creates a need for the development of molecular markers to investigate the epidemiology of such infections. To be useful, these methods have to be carefully evaluated to ascertain their performance. Using the guidelines proposed by the European Study Group on Epidemiological Markers (28), we find that the application of microsatellite polymorphism to S. cerevisiae is a powerful method. The development of such molecular markers is important for a better understanding of Saccharomyces infections. For example, in case of relapse, the method could be used to trace isolates, to test whether the infection is due to the persistence of the microorganism or reinfection with a new strain. Also, since the risk of transmission of S. cerevisiae from patient to patient has been suggested (30), it is important to identify isolates in order to demonstrate the origin of the infection and further apply appropriate prevention measures. This is particularly true for S. boulardii, which is increasingly reported as responsible for bloodstream infection and for which contamination of catheters related to hand carriage has been hypothesized (13). In the present work, reproducibility of the method and stability of the genomic regions were confirmed. To our knowledge, the level of discrimination between strains obtained with this method has never been previously reached with other methods applied to S. cerevisiae typing, confirming the high degree of polymorphism of these regions among strains of S. cerevisiae. All clinical strains collected from patients not treated with S. boulardii exhibited a distinct pattern, with the exception of isolates collected from the same patients and a cluster of four isolates collected from French patients that shared their pattern with that of a commercial baker yeast. In this case it may be postulated that the colonization of the patients results from the use of this strain in the diet, as has already been shown in another study (23).

Another important advantage of the method is its portability. Since all results can be expressed as a number of repeats, computer translation is easy, offering the opportunity to compare results from different groups via the Internet. This is an important issue for further comparisons of larger sets of strains isolated in very different settings, e.g., human medicine, environmental studies, and wine production. A similar approach involving the multilocus sequence typing method applied to bacteria has been undertaken (8, 9). Also it should be underlined that the technique can be used with a nonradioactive method, as has already been done for A. fumigatus (3).

The method also appears to be specific to the S. cerevisiae species since isolates belonging to the closely related species S. pastorianus, S. paradoxus, and S. bayanus failed to be reproducibly amplified. This must be compared to results obtained for C. albicans (10), where Candida krusei and Candida parapsilosis failed to be amplified. This could be due to the frequent mutations observed in the DNA sequences flanking repeats, leading to negative PCR results (25). The positive PCR results obtained with S. boulardii isolates strongly suggest that S. boulardii is a strain of S. cerevisiae rather than a different species. Until now, it has not been possible to identify S. boulardii in a routine manner. Phenotypic traits such as the lack of galactose assimilation and the lack of alpha-glucosidase activity are characteristics of S. boulardii (19). However none of these traits allows a definitive identification, and confirmation requires multiple enzymatic restrictions of mitochondrial DNA (M. Maillé, P. V. Nguyen, S. Bertout, and J. Bastide, Program abstr. Int. Soc. Hum. Anim. Mycol., abstr. P511, 2000). Interestingly, even the highly discriminatory randomly amplified polymorphis DNA method failed to differentiate between isolates of S. cerevisiae and S. boulardii (21). In contrast, we found that sequence (CAG)9 at locus 4 is specific for this strain and thus constitutes a rapid alternative for an accurate identification of S. boulardii. These data allow us to confirm that Saccharomyces fungemias diagnosed in our patients treated with S. boulardii therapy (eight cases) are indeed due to this particular strain. The lack of polymorphism among S. boulardii isolates is in accordance with results obtained with EcoRI-generated RFLP, while virulence studies with an animal model demonstrated significant differences among isolates (17).

Considering the small number of clinical isolates different from S. boulardii, it was not possible to test the hypothesis of a nonrandom distribution depending on the site of isolation. The fact that clinical isolates did not cluster argues against the hypothesis of common virulent traits characteristic for these isolates. This may suggest that any thermotolerant (viability at 37°C) Saccharomyces strain is able to transiently colonize the human gastrointestinal tract. However, this idea conflicts with previous reports demonstrating that virulence as tested in a murine model is more frequent in clinical isolates and is associated with particular patterns generated with an RFLP typing method (5, 6). The gathering of laboratory strains supports the opinion of Mortimer, i.e., that yeast molecular biology studies were mostly carried out on a very limited panel of strains, possibly derived from Lindegren's original strain (22). This factor must be taken into consideration in further medical studies.

The close relationship between most French wine strains may be due either to the use of identical starter strains to initiate the fermentation process or to the clonal reproduction of natural strains originating from the same geographic area. The latter possibility is supported by the clustering of Asian isolates. Similarly, strains of the yeast Cryptococcus neoformans var. gatti, which have an environmental ecological niche, show geographical clustering, suggesting clonal reproduction (4).

In conclusion, the analysis of microsatellite polymorphism is a reliable method for S. cerevisiae strain identification, including S. boulardii, which is an increasing cause of fungemia in hospitalized patients. The phylogenetic data obtained with this method will be useful in further studies in the field of population genetics.