Glucosylceramide Contained in Koji Mold-Cultured Cereal Confers Membrane and Flavor Modification and Stress Tolerance to Saccharomyces cerevisiae during Coculture Fermentation

Abstract

In nature, different microorganisms create communities through their physiochemical and metabolic interactions. Many fermenting microbes, such as yeasts, lactic acid bacteria, and acetic acid bacteria, secrete acidic substances and grow faster at acidic pH values. However, on the surface of cereals, the pH is neutral to alkaline. Therefore, in order to grow on cereals, microbes must adapt to the alkaline environment at the initial stage of colonization; such adaptations are also crucial for industrial fermentation. Here, we show that the yeast Saccharomyces cerevisiae, which is incapable of synthesizing glucosylceramide (GlcCer), adapted to alkaline conditions after exposure to GlcCer from koji cereal cultured with Aspergillus kawachii. We also show that various species of GlcCer derived from different plants and fungi similarly conferred alkali tolerance to yeast. Although exogenous ceramide also enhanced the alkali tolerance of yeast, no discernible degradation of GlcCer to ceramide was observed in the yeast culture, suggesting that exogenous GlcCer itself exerted the activity. Exogenous GlcCer also increased ethanol tolerance and modified the flavor profile of the yeast cells by altering the membrane properties. These results indicate that GlcCer from A. kawachii modifies the physiology of the yeast S. cerevisiae and demonstrate a new mechanism for cooperation between microbes in food fermentation.

INTRODUCTION

In nature, microbial communities are formed through metabolic and physiochemical interactions among different microorganisms (1). The pH of cereals that have ripened and dropped from the husk to the ground is neutral to alkaline (2–4). Fermentation microbes, including Saccharomyces cerevisiae, Lactobacillus lactis, and Acetobacter aceti, efficiently utilize carbohydrates in the cereals to produce organic acids (1). As a result, the microbial communities that colonize such carbohydrate-rich cereals, as well as those found in fermented foods, reduce the pH to 4.0 to 6.0. After these microbes establish an acidic pH, they can dominate the environment because they can grow faster at acidic pH values than other microbes. Therefore, these microbes must first adapt to the alkaline environment at the initial stage of colonization on cereal; such adaptation mechanisms are also crucial for industrial fermentation.

Traditional alcoholic beverages are produced via the fermentation of cereals by the yeast S. cerevisiae, which can efficiently catabolize glucose to produce ethanol. However, because S. cerevisiae is incapable of hydrolyzing starch to glucose, additional biological catalysts are often added to the fermentation reaction. For example, malt catalyzes starch hydrolysis in wheat- and barley-derived beverages such as whiskey and beer. In East and Southeast Asia, various cereals fermented by fungi are widely used as starch-hydrolyzing catalysts for the production of rice-derived alcoholic beverages. These include Japanese sake (5, 6) and shochu, Korean Makgeolli, and Chinese Huangjiu, as well as various other fermented foods such as soy sauce and miso (7). Koji is a mixture of steamed cereals cultured with nonpathogenic fungi, such as Aspergillus oryzae and Aspergillus luchuensis, including Aspergillus kawachii and Aspergillus awamori (the national fungi of Japan). Koji serves as a source of the enzymes that degrade cereal-derived high-molecular-weight compounds. As described above, the main role of koji is to facilitate the enzymatic conversion of polysaccharides into sugars. However, other functionally important, but as yet unidentified, interactions between yeast and koji, especially between the microbes, might also occur.

Sphingolipids are ubiquitous in eukaryotes, and their structures are highly diverse (see reference 8 and references therein). Recent studies have shown that sphingolipids are synthesized through the membrane-trafficking system and transported across membranes by various mechanisms (9–11). Many single-celled eukaryotes (i.e., fungi and protists) contain both glucosylceramide (GlcCer) and inositolphosphoceramide (IPC) and their more complex metabolites. However, although Saccharomyces kluyveri, Candida, and Cryptococcus spp., which are phylogenetically closely related to S. cerevisiae, have GlcCer, S. cerevisiae does not (12, 13). Various physiological roles have been suggested for GlcCer (reviewed in reference 8). For example, disruption of UGCG (which encodes GlcCer synthase) causes mammalian embryonic lethality (14), although UGCG loss-of-function mutations have little effect on cell growth in continuous culture (15, 16). These results suggest that GlcCer and its metabolites are indispensable for normal mammalian development. In the pathogenic fungus Cryptococcus neoformans, mutations in genes relevant to GlcCer biosynthesis cause a partial reduction in growth, especially at alkaline pH (17). Therefore, GlcCer is essential for the growth of this fungus in host extracellular environments, which are characterized by a neutral/alkaline pH (18). Nonetheless, whether GlcCer has interspecies roles when GlcCer-producing organisms live in the same environment with non-GlcCer-producing organisms is not yet known.

We hypothesized that GlcCer produced by A. kawachii on koji might affect the physiological and physicochemical characteristics of S. cerevisiae when the two microbes are cocultured for the fermentation of rice-derived alcohol. We found that A. kawachii-derived GlcCer modifies the physiological characteristics of S. cerevisiae, including its alkali and ethanol tolerance and flavor profile. GlcCer-induced tolerance resulted from changes in the membrane properties. These novel insights demonstrate a new mechanism of cooperation between microbes in food fermentation and a new technical approach for the modification of fermentation.

MATERIALS AND METHODS

Materials.

All chemicals used were of analytical grade. Ceramide, phytosphingosine, and GlcCers from soybean, Grifola frondosa, and wheat were purchased from Funakoshi Co., Ltd., Tokyo, Japan. The S. cerevisiae strains S-2 shochu yeast (a shochu industrial strain) (19) and BY4743 yeast (a laboratory strain [MATa/α his3Δ1/his3Δ1 leu2Δ0/leu2Δ0 MET15/met15Δ0 LYS2/lys2Δ0 ura3Δ0/ura3Δ0]) were obtained from the Brewing Society of Japan (Tokyo, Japan) and EUROSCARF (Frankfurt, Germany), respectively. A. kawachii (20) was obtained from Higuchi Moyashi Co., Ltd. (Osaka, Japan). The koji samples were kind gifts from Tenzan Sake Brewery (Saga, Japan).

Extraction of lipids from koji for alkali tolerance testing.

Ten grams of pregelatinized rice or koji (Tokushima Seiko, Tokushima, Japan) was mixed with 30 ml of chloroform-methanol (2:1 [vol/vol]) and sonicated for 5 min. The liquid fraction was recovered after centrifugation at 1,200 × g for 5 min. The liquid was evaporated to dryness under vacuum and dissolved in 7.5 ml of ethanol. The resulting solution was used as the lipid-containing ethanol fraction.

Purification of GlcCer.

GlcCers were purified as previously described (21, 22). Briefly, the extracted lipids were dried in a centrifugal evaporator and dissolved in 4 ml of hexane. The sample was separated by silica gel chromatography (56 cm³) eluted with ethyl acetate-methanol (9:1 [vol/vol]). The collected fractions containing GlcCer were dried and dissolved in 4.5 ml of chloroform-methanol (2:1 [vol/vol]). Four milliliters of each sample was injected into the 500-μl injection loop of a high-pressure liquid chromatography (HPLC) system. The parameters for HPLC purification of sphingolipids were as follows: column, Inertsil SIL 100 A, 5 μm, 4.6 mm by 250 mm (GL Science, Tokyo, Japan); mobile phase, buffer A (chloroform) and buffer B (95% methanol–5% water); gradient program, 0 min of 100% buffer A–0% buffer B, 15 min of 75% buffer A–25% buffer B, 20 to 40 min of 10% buffer A–90% buffer B; and flow rate, 0.7 ml/min. Fractions were collected at 0.5- or 1-min intervals and used for thin-layer chromatography (TLC) analysis, which was followed by staining with 2 mg/ml orcinol-containing 70% (vol/vol) sulfuric acid.

Extraction and fractionation of lipids from yeast and koji.

Lipids were extracted and fractionated as previously described (21, 22). Briefly, the freeze-dried BY4743 yeast or pregelatinized koji (rice fermented with A. kawachii, 0.3 g) was extracted with 2.0 ml of chloroform-methanol (1:1 [vol/vol]). The extracted lipid was mixed with 2.0 ml of 0.8 M KOH-methanol, followed by incubation at 42°C for 30 min. The samples were mixed with 5.0 ml of chloroform and 2.5 ml of water and centrifuged at 1,200 × g for 5 min. The organic phase was collected and used for further analyses. The amount of GlcCer was calculated based on a standard curve of the signal intensities of TLC spots generated using a standard material (cerebroside; Matreya, Pleasant Gap, PA) with ImageJ software.

TLC analysis of lipids from yeast cultured with GlcCer.

Extracted lipid fractions were evaporated to dryness and dissolved in 150 μl of chloroform-methanol (2:1 [vol/vol]). A 40-μl aliquot was spotted onto a silica gel TLC plate (Merck Silica Gel 60; Millipore, Billerica, MA). The plate was dried and developed with chloroform-methanol-acetic acid-water (20:3.5:2.3:0.7 [vol/vol/vol/vol]) in a presaturated TLC chamber. The spots were detected by spraying the plate with 50% (vol/vol) sulfuric acid and heating the plate at 120°C for 1 h.

Electrospray ionization tandem mass spectrometry (ESI-MS/MS).

Mass spectrometric analysis of GlcCer was performed as previously described (21, 22). Briefly, extracted lipids purified as described above were dissolved in 50 μl of chloroform-methanol (1:1 [vol/vol]). Methanol (950 μl) was added to the samples, and the samples were infused into an ion trap mass spectrometer (HC Ultra; Bruker Daltonics, Bremen, Germany) through an on-line syringe pump (flow rate, 3 μl/min). Nitrogen (at 4 liter/min, 10 lb/in², and 300°C) was used for desolvation and as a nebulizer gas. Ions were detected in positive ion mode, with a capillary voltage of 4 kV and an end plate offset of 0.5 kV. For spectrum acquisition, a mass range of 50 to 1,500 m/z and a scan speed of 4,000 m/z per s were adopted. Helium gas was used for collision-induced dissociation. The isolation width was 4 m/z. Multiple-stage sequencing of MS/MS was performed using a fragmentation amplitude of 1.0 V, ramped from 30 to 200% within 40 ms for each spectrum, and a fragmentation cutoff of 27% of the precursor ion m/z.

Alkali tolerance assay with or without lipids extracted from koji.

Liquid culture medium (1.5 ml) containing 0.67% (wt/vol) Difco yeast nitrogen base without amino acids, a 790-mg/liter complete supplement mixture, 2% (wt/vol) glucose, and 1% (vol/vol) ethanol or lipid-containing ethanol was prepared and adjusted to pH 6.5 or 8.0. S. cerevisiae shochu yeast cells were inoculated into the culture medium at an optical density at 600 nm (OD₆₀₀) of 0.1 (10⁶ cells/ml). The culture was incubated at 30°C with shaking, and the OD₆₀₀ was measured. The culture experiments were performed in triplicate with independent cultures.

Alkali tolerance assay with or without added GlcCers.

Synthetic medium containing 0.67% (wt/vol) Difco yeast nitrogen base without amino acids, a 790-mg/liter complete supplement mixture, 2% (wt/vol) glucose, 0.0015% (vol/vol) NP-40, and 1% (vol/vol) ethanol with or without 40 μg of GlcCer/ml was adjusted to pH 6.5 or 8.0 and inoculated with BY4743 yeast cells at an OD₆₀₀ of 0.1. The culture volume was 0.3 ml for GlcCer purified from koji and 1.5 ml for other GlcCers. The culture was incubated at 30°C for 17 h and then mixed, and the OD₆₀₀ was measured. The culture experiments were performed in triplicate with independent cultures. To investigate the time course of yeast growth in the presence of various GlcCers, the OD₆₀₀ of independent duplicate cultures was measured. The maximum specific growth rate (μ_max) was calculated according to the method described previously (23–26).

Culturing A. kawachii.

For liquid cultures, A. kawachii (20) mycelium powder (5 mg) was inoculated into 200 ml of Difco potato dextrose broth (24 g/liter; Becton Dickinson, Sparks, MD). The culture was incubated at 30°C with shaking for 3 days. The culture was centrifuged, and the cells were washed with sterile-distilled water. For solid cultures, 30% (wt/wt) water was added to the rice, which was then soaked for 1 h and autoclaved at 121°C for 15 min. Thereafter, 5 mg of A. kawachii mycelium powder was added, mixed, and incubated at 30°C for 3 days. Samples were lyophilized and frozen for further analyses.

Measurement of yeast membrane properties using time-resolved fluorescence spectroscopy.

The properties of the membrane of living yeast cells were analyzed as described previously (27). BY4743 yeast cells (OD₆₀₀ of 0.1) were inoculated into synthetic medium containing the solvent (1% [vol/vol] ethanol), 0.0015% (vol/vol) NP-40, and 40 μg of GlcCer/ml purified from soybean and incubated for 17 h without shaking. Cells were collected by centrifugation, washed twice with TE buffer (10 mM Tris-Cl, 1 mM EDTA [pH 8.0]), and labeled with 0.5 μM TMA-DPH (1-[4-trimethylamino]-phenyl]-6-phenyl-1,3,5-hexatriene; Invitrogen, Carlsbad, CA) at 25°C in the dark for 10 min. The cells were washed twice, resuspended in TE buffer at 0.25 OD₆₀₀, and placed on ice until use. The cell suspensions were warmed at 25°C for 3 min before measurement. A FluoroCube (Horiba, Kyoto, Japan) was used to measure the time-resolved fluorescence anisotropy of TMA-DPH. The instrument was equipped with a polarizing device and a 375-nm laser diode (NanoLED 375L; Horiba) operated with a pulse frequency of 1 MHz. TMA-DPH-labeled cells were placed in a quartz cuvette at an OD₆₀₀ of 0.4. Fluorescence emission was measured at 460 nm and 25°C. The data were analyzed using DAS6 decay analysis software (version 6.3; Horiba). The simplest model for the restricted motion of the fluorochromes in the membrane, based on Brownian diffusion of the label in a cone with a wobbling diffusion constant, led to the following single exponential approximation of the anisotropy decay with time, r(t): r(t) = (r₀ − r_∞) × exp(−t/θ) + r_∞, where r₀ is the maximum anisotropy, r_∞ is the limiting anisotropy, and θ is the rotational correlation time in nanoseconds. To obtain structural information for the membrane, the order parameter (S) was calculated using the following equation: S = (r_∞/r₀)^1/2.

To determine the dynamic nature of the membrane, the rotational diffusion coefficient (D_w) was calculated using the following equation: D_w = (r₀ − r_∞)/6θr₀, where τ represents the fluorescence lifetime. Experiments were performed in triplicate with independent cultures.

Ethanol tolerance assay with or without added GlcCers.

Synthetic medium containing 0.67% (wt/vol) Difco yeast nitrogen base without amino acids, a 790-mg/liter complete supplement mixture, 2% (wt/vol) glucose, 0.0015% (vol/vol) NP-40, and 8% (vol/vol) ethanol with or without 40 μg of soybean GlcCer/ml was inoculated with BY4743 yeast cells at an OD₆₀₀ of 0.1. The culture size was 1.5 ml. The culture was incubated at 30°C for 17 h and mixed homogenously. The OD₆₀₀ was measured. These culture experiments were performed in triplicate with independent cultures.

Alcoholic fermentation assays and flavor analysis.

Synthetic medium (5 ml) containing 0.67% (wt/vol) Difco yeast nitrogen base without amino acids, 790 mg/liter complete supplement mixture, 2% (wt/vol) glucose, 0.0015% (vol/vol) NP-40, and 1% (vol/vol) ethanol with or without 4 mg of soybean GlcCer/ml was inoculated with industrial shochu yeast cells at an OD₆₀₀ of 0.1, followed by incubation for 7 days at 15°C with liquid paraffin on top of the culture. The fermentation broth was collected after centrifugation. Volatile fermentation products produced during fermentation were analyzed using headspace gas chromatography-mass spectrometry (GC-2010, GCMS-QP2010; Shimadzu, Kyoto, Japan) with a DB-WAX column (60 m; internal diameter, 0.25 mm; 0.5 μm; Agilent Technologies, Palo Alto, CA). The carrier gas was helium, with a column headspace pressure of 125.5 kPa and a flow rate of 1.12 ml/min. The volatile fermentation product targets were ethyl acetate, n-propanol, isobutanol, isoamyl acetate, isoamyl alcohol, and ethyl caproate. For target products analysis, 900 μl of fermentation broth and 100 μl of internal standard mixture (methyl hexanoate at 5 mg/liter and n-amyl alcohol at 200 mg/liter) were placed into a 10-ml glass vial on ice. The vial containing the fermentation broth and an internal standard mixture was sealed with a magnet cap. The vial was incubated at 4°C with stirring for 10 min using an AOC-5000 autoinjector (Shimadzu). After incubation, 1,000 μl of headspace gas was injected onto the column. The gas chromatography temperature program was as follows: 40°C for 5 min, raised to 200°C at 5°C/min. The data were analyzed with GCMS solution software. Experiments were performed in triplicate with independent cultures.

Statistical design and analysis.

When analyzing microbial growth, to ensure a strict and objective comparison of sample averages, the experiments were always performed with independent cultures, the number of samples was always 10 or fewer, and P values were always calculated from the raw data with respective independent error variables. In order to minimize errors, cells were basically incubated without agitation. The differences between the averages of two groups were evaluated with unpaired one-sided Student t tests. The differences among the averages of three or more groups were evaluated using analysis of variance (ANOVA), followed by Tukey's or Dunnett's post hoc multiple-comparison test.

RESULTS

Koji-derived lipid fractions and GlcCer confer alkali tolerance to yeast.

In order to elucidate the interaction between yeast and Aspergillus spp., we formulated a testable hypothesis: yeast exposed to substances from coexisting Aspergillus spp. acquire alkali tolerance. To test this hypothesis, we first examined whether a lipid fraction extracted from koji or steamed rice affected the growth of a shochu yeast of S. cerevisiae at alkaline pH. The yeast strain showed decreased growth in alkaline liquid medium (pH 8.0) compared to the growth of yeast in neutral liquid medium (pH 6.5) (Fig. 1A and B). The growth of the yeast at pH 8.0 increased significantly when exogenous koji-derived lipids were supplied (P < 0.001; Fig. 1A). Growth also increased to a lesser extent when steamed rice-derived lipids were supplied (Fig. 1C). In contrast, supplementation with these lipid fractions only slightly affected or did not affect the growth of yeast in medium at pH 6.5 (P > 0.05; Fig. 1B).

FIG 1.

Lipids and glucosylceramide (GlcCer) extracted from koji confer alkali tolerance to yeast. (A and B) Growth profiles of yeast (Saccharomyces cerevisiae) cultured with or without lipids from koji. Ethanol alone (15 μl) or an ethanol solution (15 μl) containing lipid extracted from pregelatinized koji was added to synthetic medium (1.5 ml; adjusted to pH 8.0 or 6.5). A shochu yeast S. cerevisiae strain was inoculated into the medium (OD₆₀₀ = 0.1) and incubated at 30°C. The OD₆₀₀ was measured as described in Materials and Methods. The culture experiments were performed in triplicate with independent cultures. (A) pH 8.0. (B) pH 6.5. (C) Yeast growth in the presence of lipids prepared from rice grain or koji. BY4743 yeast S. cerevisiae cells were inoculated into medium (pH 8.0) containing lipids prepared from rice grain or koji at an OD₆₀₀ of 0.1 and incubated at 30°C. The OD₆₀₀ was measured after 17 h of culture as described in Materials and Methods. The results are the mean values with standard errors of triplicate independent experiments. The statistical significance of the differences between the averages was assessed with an unpaired one-tailed Student t test or a Tukey post hoc multiple-comparison test. *, P < 0.05; **, P < 0.01; ***, P < 0.001.

Purification of alkali tolerance-conferring lipids from koji.

To identify the lipid molecule responsible for conferring alkali tolerance, we purified the lipids from koji by silica gel chromatography (Fig. 2A) and tested the alkali tolerance-conferring activity of the separated lipid fractions. The alkali tolerance-conferring activity was mainly attributed to three lipid classes (lipids 2 to 4, Fig. 2B). Because previous studies have reported that unsaturated fatty acids (28), which we estimated to be lipid 2, and sterols (29), which we estimated to be lipid 3, affect the fermentation efficiency, we focused on lipid 4, and purified it to homogeneity by HPLC (Fig. 2C). The purified lipid did indeed confer alkali tolerance to yeast strain BY4743 (Fig. 2D). To obtain structural information, the purified lipid was subjected to mass spectrometry. Mass spectrometric analysis showed that the lipid contains several ions with m/z values of 736.6, 764.6, 776.6, and 792.6 (Fig. 2E), which corresponded to the m/z GlcCer d18:2/C16:0 h (barley or rice), d18:2/C20:0 h (barley or rice), 9-Me-d18:2/C16:0 h (fungi), and d18:2/C18:0 h (barley or rice), respectively, according to our previous studies (21, 22) and a study from another group (30). The structure of one of the ions with a m/z of 776.6 was further investigated using MS/MS and fragment ion analysis. The product ions of the precursor with a m/z of 776.6 corresponded to those of GlcCer 9-Me-d18:2/C16:0 h (m/z 614.6 is Y₀ [ceramide moiety], m/z 496.4 is O [glycosyl sphingoid base moiety], and m/z 346.3 is T [fatty acid moiety]) (Fig. 2F), which is present in Aspergillus spp., according to our previous study (21, 22). These results indicate that the GlcCer contained in koji confers alkali tolerance to yeast.

FIG 2.

Purification of an alkali tolerance-conferring lipid from koji. (A) Freeze-dried koji (50 g) was extracted with 150 ml of chloroform/methanol (2:1 [vol/vol]), filtered, evaporated to dryness, and solubilized in 200 ml of hexane. The solubilized lipid was separated by silica gel chromatography and eluted with ethyl acetate/methanol (9:1 [vol/vol]). The eluate was collected in 10-ml fractions. Small aliquots (1:15 [vol/vol]) of the fractions were applied to a Merck Silicagel60 Plate (Merck, Schuchardt, Germany), developed with chloroform-methanol-acetate-water (20:3.5:2.3:0.7 [vol/vol/vol/vol]), and sprayed with 2-mg/ml orcinol-containing 70% (vol/vol) sulfuric acid. Lipid fractions 1 to 4 (shown on the TLC plate) were recovered. (B) The lipids were mixed in synthetic medium at 1% (vol/vol), and BY4743 yeast cells were inoculated into the medium at a density of 10⁶ cells/ml. The culture was incubated at 30°C for 17 h, and the OD₆₀₀ of the culture was measured. (C) HPLC purification profile of lipid 4. The underlined fraction, corresponding to GlcCer, was recovered. (D) The alkali tolerance-conferring activity of the purified lipid. (E) Positive ESI-MS spectrum of the purified lipid in panel C. The purified lipid was applied to an ion trap mass spectrometer (HC Ultra; Bruker Daltonics) with an electrospray ion source. (F) Positive ESI-MS/MS spectrum of product ions generated from the precursor ion (m/z 776.6) in panel E. Detailed methods are presented in Materials and Methods. The statistical significance of the difference between the averages was assessed using the unpaired one-tailed Student t test (**, P < 0.01).

GlcCers from various organisms confer alkali tolerance to yeast.

The GlcCer purified from koji (barley koji) contains a mixture of GlcCer molecules from A. kawachii and barley. To determine which structural features of the GlcCer confer alkali tolerance to yeast, we used GlcCers from three different species: soybean, Grifola frondosa, and wheat. GlcCers from soybean and barley have the same structure (d18:2^4t.8t/C16:0 h; Fig. 3B), while GlcCer from G. frondosa has a 9-Med18:2^4t.8t/C16:0 h structure (Fig. 3C) (30). The sphingoid base of GlcCer from G. frondosa is shared with that of GlcCer from A. kawachii (9-Me-d18:2^4t.8t/C18:1 h) (Fig. 3C). GlcCer from wheat (d18:1^8c/C16:0 h) is different from the GlcCers in the five other organisms we examined here (A. kawachii, rice, barley, soybean, and G. frondosa) (Fig. 3D). From our observations of the cultures, it seemed clear that GlcCer affected the growth of BY4743 yeast because GlcCer-supplemented yeast cultures had a rather dispersed pellet compared to control yeast (Fig. 3A). Next, in order to quantitate the effect of the various GlcCer species on alkali tolerance, we performed a time course analysis of the growth of yeast in the presence or absence of these GlcCers. GlcCers from soybean, G. frondosa, and wheat conferred alkali tolerance to BY4743 yeast in a time-dependent manner (Fig. 3B to D). The maximum specific growth rates (μ_max) were significantly increased in cells cultured with all GlcCers (P < 0.05; Fig. 3E), with soybean GlcCer having the largest effect. GlcCer did not significantly (P > 0.05) increase yeast growth at pH 6.5 (OD₆₀₀ values after 17 h of culture: control, 3.96 ± 0.033; 40 μg of G. frondosa GlcCer/ml, 4.05 ± 0.045; 40 μg of soybean GlcCer/ml, 4.04 ± 0.072), indicating that GlcCer did not function as a nutrient in yeast. In order to rule out the possibility that the alkali tolerance-conferring effect was due to the small size of the culture, we increased the culture size from 1.5 to 5 ml, and the effect was again statistically significant (P < 0.05; Fig. 3F). The finding that all of the tested GlcCers conferred alkali tolerance to the yeast indicates the broad structural specificity of the alkali tolerance-conferring activity.

FIG 3.

GlcCers from various sources confer alkali tolerance to yeast. (A) Growth of S. cerevisiae BY4743 yeast cells in the presence of GlcCers. S. cerevisiae BY4743 yeast cells were inoculated at an OD₆₀₀ of 0.1 into 1.5 ml of synthetic medium (pH 8.0) containing 40 μg of wheat GlcCer/ml, followed by incubation at 30°C for 17 h. (B to D) Growth profile of BY4743 yeast cells cultured in 1.5 ml of synthetic medium containing 40 μg of GlcCer/ml from soybean (B), G. frondosa (C), and wheat (D). Yeast cells were inoculated into the media at a density of 10⁶ cells/ml, followed by incubation at 30°C. The OD₆₀₀ of the culture was then measured. The results are the mean values with standard errors of duplicate independent experiments. The statistical significance of the difference between the averages was assessed using an unpaired one-tailed Student t test (**, P < 0.01; *, P < 0.05). (E) μ_max (maximum specific growth rate, h⁻¹) of BY4743 yeast cultured with various GlcCers. The statistical significance of the differences between the averages was assessed with ANOVA followed by a post hoc Dunnett multiple-comparison test (*, P < 0.05; **, P < 0.01). (F) Growth of yeast cells in a 5-ml culture with added GlcCer. BY4743 yeast cells were inoculated at an OD₆₀₀ of 0.1 into synthetic medium (pH 8.0) containing 40 μg of soybean GlcCer/ml and incubated at 30°C for 17 h. The OD₆₀₀ of the culture was then measured. The results are the mean values with standard errors of triplicate independent experiments. The statistical significance of the differences between the averages was assessed with the unpaired one-tailed Student t test (*, P < 0.05; **, P < 0.01). Detailed methods are presented in Materials and Methods.

GlcCer, but not its degraded forms, mainly confers alkali tolerance.

To determine whether degraded forms of GlcCer, such as ceramide and sphingosine bases, confer alkali tolerance, we first analyzed the TLC profile of lipids extracted from BY4743 yeast cultured with GlcCer (40 μg/ml). We found that free ceramides and sphingoid bases were not detectable in the yeast cell-associated lipid fraction (Fig. 4A). We then examined the possibility that small amounts of ceramides or sphingoid bases, too low to be detected by the TLC staining method, could confer alkali tolerance. We tested whether supplementation with exogenous GlcCer, ceramide, or phytosphingosine conferred alkali tolerance to BY4743 yeast in a time-dependent manner (Fig. 5). Interestingly, while a significant effect was observed at a lower concentration of ceramide (4 μg/ml, corresponding to 1/10 the concentration of added GlcCer), no significant or little effect was observed at 1.3 μg/ml (corresponding to 1/30 the concentration of added GlcCer, P > 0.05) (Fig. 5B). Exogenous phytosphingosine (1.3 to 40 μg/ml) seemed to be toxic to the yeast cells (P < 0.05; Fig. 5C). Semiquantification of lipids with sulfuric acid staining demonstrated that the amount of ceramide in the total lipid fraction extracted from yeast cells cultured with GlcCer supplementation contained no more than 1/30 the amount of GlcCer (Fig. 4B). From these results, we concluded that the alkali tolerance acquired by yeast in coculture with GlcCer was primarily conferred by GlcCer, although undetectable degraded products might have made minor contributions to the total alkali tolerance.

FIG 4.

Metabolic fate of GlcCer in yeast. Total lipid profile of yeast cells treated with GlcCer. The total lipids of S. cerevisiae BY4743 yeast cells cultured with or without GlcCer (in 5 ml of synthetic medium adjusted to pH 8.0) for 17 h were extracted with chloroform-methanol (2:1 [vol/vol]), separated by TLC, and visualized with 50% (vol/vol) sulfuric acid as described in Materials and Methods. In the panels, S and L represent small and large amounts of each sample set, respectively. (A) Sc S represents 2.4 × 10⁷ S. cerevisiae cells, Sc L represents 7.2 × 10⁷ S. cerevisiae cells, “Sc + Soybean GlcCer S” represents 2.4 × 10⁷ S. cerevisiae cells with 40 μg of soybean GlcCer/ml, “Sc + Soybean GlcCer L” represents 7.2 × 10⁷ S. cerevisiae cells with 40 μg of soybean GlcCer/ml, “Soybean GlcCer S” represents 16 μg of soybean GlcCer, “Soybean GlcCer L” represents 24 μg of soybean GlcCer, “Bovine cerebroside S” represents 16 μg of bovine cerebroside, “Bovine cerebroside L” represents 24 μg of bovine cerebroside, “Ceramide S” represents 15 μg of ceramide, “Ceramide L” represents 23 μg of ceramide, “Phytosphingosine S” represents 24 μg of phytosphingosine, and “Phytosphingosine L” represents 32 μg of phytosphingosine. (B) Ceramides 1, 2, 3, 4, and 5 represent 0.23, 0.77, 2.3, 7.7, and 23 μg of ceramide, respectively. The results shown are representative of two to four independent experiments.

FIG 5.

Dependence of the alkali tolerance-conferring activity on the sphingolipid concentration. S. cerevisiae BY4743 yeast cells were cultured in the presence of GlcCer from soybean (A), ceramide (B), and phytosphingosine (C). The cells were inoculated at an OD₆₀₀ of 0.1 into synthetic medium (pH 8.0) containing different concentrations of sphingolipids and incubated at 30°C. The OD₆₀₀ of the culture was then measured. The results are the mean OD₆₀₀ values with standard errors of duplicate independent cultures. The statistical significance of the differences between the averages was assessed with ANOVA followed by a post hoc Tukey multiple-comparison test (different letters indicate statistically different values at P < 0.05). Detailed methods are described in Materials and Methods.

Koji contains abundant GlcCer, and yeast is exposed to a high concentration of GlcCer during food fermentation.

Next, we investigated whether this phenomenon also occurs in a food fermentation environment. Interestingly, the GlcCer content of rice fermented with A. kawachii was found to be much higher than the GlcCer content of unfermented rice (Fig. 6A). Furthermore, a fermentation mash containing rice fermented with A. oryzae (Fig. 6B) or A. kawachii (31) contained abundant GlcCer. These findings supported the idea that koji could serve as a major GlcCer supplier for yeast in coculture fermentation. The GlcCer content varied among koji samples (Fig. 6C), suggesting that manufacturing manipulations affect the GlcCer content of koji. These results indicate that under food fermentation conditions, the GlcCer content of the koji is a determinant of fermentation.

FIG 6.

Effect of GlcCer on yeast under the fermentation conditions. (A) Lipid profiles of rice and rice fermented with A. kawachii. Total lipids from rice and rice fermented with A. kawachii were extracted with chloroform-methanol (2:1 [vol/vol]), separated by TLC, and visualized with 50% (vol/vol) sulfuric acid. (B) Lipid profile of the fermentation mash. Lanes 1 to 3, 20, 40, and 60 mg of fermentation mash, respectively; lane 4, ceramide (16 μg); lane 5, GlcCer from soybeans (15 μg). Fermentation mash containing yeast, rice, and koji A. oryzae was filtered by centrifugation, and the total lipids in the solid portion of the mash were extracted with chloroform-methanol (2:1 [vol/vol]), separated by TLC, and sprayed with 50% (vol/vol) sulfuric acid for visualization, as described in Materials and Methods. (C) GlcCer content of koji prepared at various rice polishing ratios. The rice polishing ratio indicates the percentage of the grain weight that remains after polishing to remove the hull and bran. The results are the mean values with standard errors of triplicate measurements of one to six independent koji samples. (D) Ethanol tolerance-conferring activity of GlcCer. S. cerevisiae BY4743 yeast cells were inoculated at an OD₆₀₀ of 0.1 into 1.5 ml of synthetic medium (pH 8.0) containing 8% (vol/vol) ethanol with or without 40 μg of soybean GlcCer/ml, followed by incubation at 30°C. The OD₆₀₀ of the culture was then measured. The results are the mean values with standard errors of triplicate independent experiments. The statistical significance of the difference between the averages was assessed using the unpaired one-tailed Student t test (*, P < 0.05; **, P < 0.01; ***, P < 0.001).

GlcCer modifies the ethanol tolerance and flavor profile of yeast.

During the manufacture of alcoholic beverages, yeast is exposed to a high concentration of ethanol, and the growth of yeast in ethanol-containing medium reflects one aspect of ethanol tolerance. Therefore, we investigated the effect of GlcCer on the growth of yeast in ethanol-containing medium. Addition of GlcCer significantly (P < 0.05) increased the growth yield of BY4743 yeast in synthetic medium containing 8% (vol/vol) ethanol in a time-dependent manner (P < 0.01; Fig. 6D). We also determined whether GlcCer affected the metabolite profile of yeast. Here, we focused on flavor-relevant metabolites in an industrial shochu yeast strain (S-2), because modification of flavor by koji GlcCer might become a crucial characteristic of the beverage. Medium from yeast fermentation with GlcCer contained significantly higher levels of ethyl acetate and isoamyl acetate and lower levels of n-propanol than the control fermentation medium (P < 0.05; Table 1). These results indicate that GlcCer also modifies the ethanol tolerance of yeast and the flavor compounds produced by the yeast.

TABLE 1.

Flavor profile of fermentation broth with or without added GlcCer^a

Flavor	Mean concn (μg/ml) ± SE		P
	Control	GlcCer treated
Ethyl acetate	2.10 ± 0.18	3.48 ± 0.18	<0.01
n-Propanol	17.24 ± 0.53	16.27 ± 0.62	<0.05
Isobutanol	51.50 ± 5.72	56.95 ± 4.92
Isoamyl acetate	0.00 ± 0.00	0.15 ± 0.13	<0.05
Isoamyl alcohol	76.46 ± 4.55	82.80 ± 6.79
Ethyl caproate	0.08 ± 0.04	0.09 ± 0.07

^aGlcCer from soybean (40 μg/ml) was added to the fermentation broth. The fermentation broth was inoculated with industrial shochu yeast S-2 and incubated at 15°C for 7 days. Fermentation tests and analyses were performed in triplicate with independent cultures. Further details are given in Materials and Methods.

The membrane of yeast cells cultured in the presence of GlcCer has altered properties.

We investigated the mechanism underlying the acquisition of the physiological changes in yeast cultured in the presence of GlcCer. Previous studies have suggested that alkali and ethanol tolerance and flavor production in yeast cells are governed by membrane properties (29, 32–35). Thus, we analyzed various membrane properties of BY4743 yeast treated with GlcCer using time-resolved fluorescence anisotropy (36, 37). Measurement of membrane rigidity with fluorescence depolarization techniques provides a sensitive, reliable, and objective method to assess the status of the membrane. One such technique uses TMA-DPH (38, 39), a molecule with a lipophilic moiety and a cationic moiety, which reflects the interfacial property of the membrane. A low value for the order parameter (S) and a high value for the rotational diffusion coefficient (D_w) reflect membrane disorder and increased rotational lipid acyl chain motion, respectively. Membranes prepared from control cells had an S value of 0.949 ± 0.002 and a D_w value of 2.5 ± 0.5 μs⁻¹ (Table 2), while membranes treated with GlcCer tended to show a modest decrease in S (0.941 ± 0.004) and a concomitant increase in D_w (3.1 ± 0.9 μs⁻¹). However, these differences were not statistically significant (S, P = 0.14; D_w, P = 0.33), possibly because the differences between the two averages were not large enough compared to the deviation caused by the technique. The decrease in the fluorescence lifetime of TMA-DPH reflects an increase in the degree of water molecule penetration into the membrane, which increases its dielectric constant. The average fluorescence lifetime (τ) of TMA-DPH in control cells was 11.9 ± 0.05 ns, while that in GlcCer-treated cells was significantly (P < 0.05) lower (11.6 ± 0.04 ns; Table 2), suggesting that more water molecules penetrated the lipid bilayer of cells treated with GlcCer. Collectively, these results suggest that treating yeast cells with GlcCer disorders the plasma membrane and accelerates the rotational motion of the acyl chains.

TABLE 2.

Effects of GlcCer on the anisotropy decay of TMA-DPH in yeast cells

Sample type	Mean ± SEMa
	τ (ns)	r0	r∞	S	Dw (μs−1)
Vehicle control	11.9 ± 0.05	0.287 ± 0.001	0.258 ± 0.001	0.949 ± 0.002	2.5 ± 0.5
GlcCer treated	11.6 ± 0.04**	0.288 ± 0.006	0.255 ± 0.003	0.941 ± 0.004*	3.1 ± 0.9

^aData are mean values of three independent culture experiments: τ, fluorescence lifetime; r₀, maximum anisotropy; r_∞, limiting anisotropy; S, order parameter; D_w, rotational diffusion coefficient. GlcCer from soybeans (40 μg/ml, final concentration) was added to the BY4743 yeast cell cultures. Cells were incubated for 17 h and then labeled with TMA-DPH. Asterisks indicate statistical significance with respect to the control: *, P < 0.15; **, P < 0.05.

DISCUSSION

Compared to unfermented cereals, such as rice and barley, the mold-cultured cereal koji contains a greater amount of GlcCer (Fig. 6A). In the present study, we showed that koji GlcCer modified several physiological properties of yeast, including stress tolerance and flavor production, and affected the fermentation profile, although the effect was modest. Making koji is a complex, laborious process, and the factors that affect subsequent fermentation have not been fully elucidated. The GlcCer content of koji varies depending on several brewing practices, such as rice polishing, and other, as-yet unknown factors (Fig. 6C). Given that yeast cells are exposed to a high concentration of GlcCer in the fermentation mash and that GlcCer induces several physiological changes in yeast, we propose that the GlcCer in koji, in addition to other lipid factors such as unsaturated fatty acids and sterols, is critical for ensuring the quality of the alcoholic fermentation.

Both the internal synthesis and external addition of unsaturated fatty acids during the alcoholic fermentation of wine, beer, and sake have been shown to play essential roles in maintaining membrane integrity and function, which influences stress adaptation and modifies the fermentation profiles (33) and final flavor (35). For example, incorporation of a gene encoding a polyunsaturated fatty acid synthase (Arabidopsis thaliana delta-12 fatty acid desaturase [FAD2]) increases the alkali and ethanol tolerance of yeast (28, 34). Furthermore, the addition of unsaturated fatty acids decreases the fluorescence lifetime (τ) of the membrane (37), and the internal synthesis or external addition of sterols enhances the ethanol and heat tolerance of yeast (29, 40). In the present study, we showed that the addition of GlcCer conferred alkali and ethanol tolerance to yeast, modified flavor formation, and decreased the fluorescence lifetime of TMA-DPH. Thus, GlcCer seemed to improve the physicochemical properties of the membrane in a manner similar to that of unsaturated fatty acids and sterols.

The “fluid mosaic” model of the biological membrane proposed that the membrane is a homogeneous lipid bilayer. However, the heterogeneous status of the membrane is now generally accepted. The heterogeneity of membranes is derived from the liquid and crystalline phases of the lipid bilayers, which are greatly affected by the lipid molecules. Indeed, sphingolipids interact with sterols through van der Waals interactions and form specific membrane regions called “rafts” (41). The physicochemical status of the membrane is also affected by the length of the fatty acids through the lateral packing of lipid molecules (42–45). The length of the fatty acids in GlcCers from plants and fungi is C₁₆ to C₂₀ (21, 22, 30), whereas the length of those in the phosphodiester-linked complex sphingolipids of yeast is usually C₂₆ (46, 47). Thus, the difference in acyl chain length between phosphodiester-linked complex sphingolipids and GlcCer might account for the increased penetration of water into the membrane (as reflected by the τ value) of the GlcCer-treated cells (Fig. 7). A scenario in which GlcCer acts as a breaker of putative “C₂₆-sphingolipid-rafts” in the plasma membrane of S. cerevisiae cells would explain why exogenous C₁₈-ceramide conferred alkali tolerance (Fig. 4B). Thus, when incorporated into S. cerevisiae, C₁₈-ceramide might also behave as a C₂₆-sphingolipid-raft breaker. Consistent with this hypothesis, ceramide, which has a shorter (C₁₆) chain length, had a larger anisotropy decay than ceramide, which has a longer (C₂₄) chain length (48). The results of the present study, together with those of preceding studies (48), indicate that exogenous GlcCer molecules are spontaneously incorporated into the outer leaflet of the lipid bilayer of the plasma membrane, which increases membrane fluidity. The change in membrane fluidity modifies various physiological properties, such as alkali and ethanol tolerance and metabolite profiles. However, it is also possible that exogenous GlcCer exerts its effects after internalization into the endomembrane compartments of the yeast cells, like other sphingolipid molecules (49–52) (Fig. 7). Whether GlcCer is incorporated into the lipid bilayer of yeast or into a specific intracellular compartment was not established in the present study. The answer awaits future technological developments.

FIG 7.

Hypothetical schematic diagram of the effect of GlcCer from mold-cultured cereal (Aspergillus kawachii koji) on yeast physiology. Yeast S. cerevisiae exposed to GlcCer from cereal molded with A. kawachii (koji), through either incorporation into the plasma membrane or internalization into intracellular endomembrane compartments, acquires ethanol and alkali tolerance and flavor modification. Exogenous GlcCer contains C₁₆ to C₂₀ fatty acids, whereas the acidic complex sphingolipids of yeast, such as IPC, MIPC, and M(IP)₂P, contain C₂₆ fatty acids. Because of the difference in the acyl chain length, the lateral packing of the lipids weakens, which induces unpacking of the membrane and decreases the fluorescence lifetime (τ) value of TMA-DPH.

The evolutionary reason why S. cerevisiae lacks the genes required for synthesizing GlcCer is unclear. Compared to the sphingolipids found in various fungal species, the acyl chain length of the major sphingolipids in S. cerevisiae is very long (typically C₂₆), and this very long acyl chain is somehow crucial for S. cerevisiae (53, 54). Interestingly, fungal GlcCer synthase exhibits strict structural specificity for the ceramide substrate, and the enzyme is likely incapable of using C₂₆-ceramide as a substrate (8, 17, 55). Therefore, during the evolution of S. cerevisiae, as GlcCer synthesis became enzymatically incompatible with the synthesis of C₂₆-sphingolipids, the ability to produce C₂₆-sphingolipids might have been selected for, even at the expense of losing GlcCer. Although S. cerevisiae is often used without other fungi for the production of various foods and beverages (e.g., bread, beer, and wine), in nature, this organism lives in the soil, where it likely forms communities with other microbes. In fact, S. cerevisiae has been shown to coexist with other fungi, such as A. oryzae, in natural environments (56). Thus, it is interesting to speculate that acquiring GlcCer from neighboring microbes might have facilitated the loss of the GlcCer synthase gene from its genome.

In conclusion, we showed here that the GlcCer contained in koji, a cereal molded with A. kawachii, modifies various physiological characteristics of S. cerevisiae: GlcCer alters its membrane properties, confers alkali and ethanol tolerance, and modifies its flavor profile. These previously unrecognized interspecies interactions demonstrate a new mechanism of cooperation between microbes in food fermentation.