Improved Phenolic Compositions and Sensory Attributes of Red Wines by Saccharomyces cerevisiae Mutant CM8 Overproducing Cell-Wall Mannoproteins

: The objective of this study was to improve the quality attributes of red wines by Saccharomyces cerevisiae (BCRC 21685) mutant CM8 with overexpression of high-mannose mannoproteins, with respective to phenolic compositions, colorimetric parameters, and consumer sensory attributes. The CM8 was mutated by ethyl methane sulfonate and showed the ability of overproducing cell wall mannoproteins selected by killer-9 toxin-containing YPD plates. Kyoho grapes were used as raw materials. It is interesting to ﬁnd that the cell wall mannoproteins isolated from CM8 mutant possessed a signiﬁcantly higher mannose content in the polysaccharide fraction (81% w / w ) than that did from parent strain (66% w / w ). The red wines made of winter grapes and CM8 (CM8-WIN) showed signiﬁcantly greater total tannins, ﬂavonols, and anthocyanins levels, as well as higher color, higher ﬂavor, and higher consumer preference than those by its SC counterpart (SC-WIN). The characteristics of the red wines studied were further elucidated by principal component analysis. Conclusively, using CM8 starter could e ﬀ ectively endow the red wine with high-quality attributes via the interactions of high-mannose mannoproteins with wine compounds.


Selection of Mannoprotein-Overexpressed Mutants
Killer 9 toxin (k9) agar plate was used to select mannoprotein-overexpressed mutants, which were resistant to k9 due to their cell walls rich in mannoprotein. Killer 9 toxin was produced by culturing Hansenula mrakii IFO 0897 (BCRC 23058) in buffered YDP medium (1% yeast extract, 2% peptone, 2% glucose, 50 mM sodium citrate buffer) [21]. After incubated at 19 • C for 14 h with 150 rpm rotation, the resultant k9-containing culture was centrifuged (9200× g, 10 min) and filtered (0.2 µm sterile filter) to give k9 supernatant. Five mL of k9 supernatant was overlayered on each of buffered YPD agar plates [19]. Serial 10-fold dilutions of each of UV or EMS-treated yeast suspensions (initially OD 600 = 0.3,~10 8 CFU/mL) were prepared with saline solution. An aliquot of the dilution (100 µL) was sampled to spread on k9-overlayered buffered YPD agar plates. After incubation at 28 • C for 48-72 h, survival colonies were selected. The candidate mutants were selected from k9 toxin plates by the following procedure. A total of 50 colonies were picked from plates for each treatment, 8 colonies among the 50 cultures were selected by second plating, and finally 2 strains (M1and M7) from UV mutants and 3 strains (CM3, CM5, and CM8) from EMS ones, which showed notable resistances against killer-9 toxin, were further compared their mannoprotein-producing ability.

Analysis of Mannose and Glucose Contents in Cell Walls
Survival colonies were enriched on buffered YPD agar plates without k9, followed by washing twice with sterilized miliQ H 2 O and collecting by centrifugation at 9200× g for 10 min at 4 • C. Cells were resuspended in 0.5 mL miliQ H 2 O containing 0.5 g of glass beads (diameter of 2 mm) and broken in a MP TM Biomedical FastPrep ® -24 Tissue Homogenizer (MP Biomedicals Europe, Eschwege, Germany) by 5 cycles of shakeup and pause for 40 s and 2 min, respectively. Cell debris was collected by centrifugation at 9,200× g for 10 min at 4 • C, washing with sterile miliQ water trice, and hydrolyzed by 200 µL of 1 M H 2 SO 4 aqueous solution at~100 • C (boiling water) for 4 h [3]. The hydrolysate was diluted with 1.8 mL miliQ water, filtered through 0.45 µm nylon membrane, and subjected to a high-performance liquid chromatography (HPLC) system. A Hitachi L-6000 HPLC system equipped with a refractive index detector IOTA 2 (Precision Instruments Inc., Des Plaines, IL, USA) and MetaCarb 87C column (300 × 7.8 mm) (Varian Inc., Palo Alto, CA, USA) was applied. The elution was miliQ water at 0.7 mL/min and 85 • C. The injection volume was 10 µL. Chromatograms were managed with Peak ABC chromatography workstation V 2.11 software. Data were measured in three replications and calibrated with mannose and glucose standard curves.

Quantification of Mannoproteins in Yeast Mutants
Mannoproteins in yeast cells were extracted according to Dikit et al. [22]. Yeast cells (dry weight 5 g, approximate 20 g wet weight) were resuspended in 20 mL phosphate buffer (pH 7.0) and autoclaved under 121 • C for 2 h. After cooling and centrifugation (9,200× g, 10 min), the supernatant was mixed with 5 volumes of 95% ethanol and settled overnight for complete precipitation. The precipitate was washed with 95% ethanol trice and freeze-dried.
Mannoproteins were collected by fast performance liquid chromatography (FPLC). A Pharmacia™ LCC-501 Plus FPLC system equipped with a UV detector (280 nm) and Affinity Chromatography Media Cellufine ® PB resin column (2.6 cm × 15 cm) (JNC Corporation, Tokyo, Japan) was applied. The column was successively eluted with adsorption buffer (0.01 M sodium phosphate, pH 7.5) for 30 min and, for another 60 min, with eluting buffer (0.01 M sodium phosphate, pH 6.7) containing gradient NaCl from 0.1 M to 0.2 M in 60 min. The flow rate was 5 mL/min. After loading sample into the column, all proteins bound onto resin during running adsorption buffer. Thereafter, mannoproteins were eluted by elution buffer with gradient 0.1-0.2 M NaCl. The eluted mannoproteins were collected, dialyzed (molecular weight cut off = 1 kDa) against deionized water to remove salts, precipitated with 5 volumes of 95% ethanol, freeze-dried, and weighted. The yield of mannoproteins was examined in three replications and on dry yeast basis.

Production of Red Wine
Kyoho grapes were washed, mechanically crashed, and pasteurized (70 • C, 10 min). Three batches of red wine were produced. For each batch, approximate 3.5 kg pasteurized grapes including pulp, peel and grapes juice (1.0 kg), which recovered from grapes crashing processes, were mixed with brown sugar (to be final 25 • Brix) in a 5-liter glass jar and inoculated with yeasts (500 mL, 1 × 10 8 CFU/mL). The mixture was pre-fermented at 18 • C for 6 h and performed primary fermentation at 28 • C for 12 d. After primary fermentation and removing the residues, wine mash was subjected to secondary fermentation at 4 • C for 14 d. After removing precipitates, the given red wine was transferred to a sterile glass jar (5 L) and incubated at 4 • C for 90 d to complete the aging process.

pH Value
A pH meter (EL20, Mettler Toledo, Schweiz) was employed to determine the pH value of wine.

Ethanol Concentration
Wine sample (10 mL) was diluted by 500 folds with deionized water and filtered through 0.45 µm nylon membrane before gas chromatography. An Agilent™ 7890A GC system (Agilent™ Inc., Palo Alto, CA, USA) equipped with an FID detector and Rtx-1301 column (30 m × 0.32 mm ID × 1.0 µm) (RESTK Corporation, Bellefonte, PA, USA) was employed. Helium was used as the carrier gas. The temperature program started from 35 • C (hold for 5 min) to 155 • C at a heating rate of 20 • C/min for column oven. Temperatures for injector and detector were set at 200 and 260 • C, respectively. The linear velocity of carrier gas was 25 cm/sec at 35 • C and a split ratio of 30:1. Isopropanol (5% v/v), as the internal standard, was added into the wine sample. One µL of sample was applied to the GC. Data were measured in three replications.

Titratable Acidity
According to the method of OIV [23], a portion of wine sample (1 mL, de-CO 2 ) was diluted with 10 mL de-CO 2 distilled water and titrated with 0.10 N NaOH aqueous solution (equivalent factor F = 0.075) to an end point of pH 8.1. The volume of NaOH solution consumed (V NaOH , mL) was record. The titratable acidity was calculated using the following equation, Equation (1), where N NaOH = the normality of NaOH solution (0.10); D = dilution fold (10), V s = sample volume (1 mL), and expressed as gram of tartaric acid equivalent (TAE) per liter of wine. Data were measured in three replications.

Total Phenolics, Flavonoids, and Tartaric Ester Contents
According to Cliff et al. [5], an aliquot of sample (0.5 mL) was diluted with 5 mL of 10% ethanol. The diluted sample was sampled (0.25 mL) and added to 0.25 mL of 0.1% HCl in 95% ethanol and 4.55 mL of 2% HCl. Each mixture was vortexed and allowed to stand for 15 min before measurement on the absorbances at 280, 320, and 360 nm for corresponding concentration of total phenolics, flavonols, and tartrate esters, respectively. Based on standard calibration curves, the total phenolics content (TPC) was expressed as mg gallic acid equivalent (GAE) per liter of wine; total flavonols content (TFC), as mg quercetin equivalent (QE) per liter of wine; and total tartrate esters content (TEC), mg caffeic acid equivalent (CAE) per liter of wine. Data were measured in three replications.
Besides, different types of anthocyanins (AC), including co-pigmented, monomeric, and polymeric ACs, were determined according to their reactivities with acetaldehyde and SO 2 [5]. A portion of wine (2 mL) was successively added with 20 µL of 20% acetaldehyde, mixed well, stayed for 45 min, and measured for the absorbance at 520 nm (noted as A ace ) in a Spectrophotometer U-1800 system (Hitachi, Tokyo, Japan). Similar procedure was done on another portion of wine (2 mL) added with 160 µL of 5% (w/v) SO 2 , giving the absorbance at 520 nm (A SO2 ). A portion of wine sample without acetaldehyde or SO 2 was directly measured the absorbance at 520 nm (A wine ). The concentrations of total, co-pigmented, polymeric, and monomeric AC were positively related to the values of A ace , (A ace −A wine ), A SO2 , and (A wine −A SO2 ), respectively [5]. Because that the absorbance is fundamentally proportional to the AC concentration, the percentages of different types of ACs in total AC can be calculated as Equations (2)-(4). All data were measured in three replications.

Total Tannin Content
Total tannin content (TTC) in wine was determined according to the method of Harbertson et al. [24]. Briefly, tannins in wine were tested based on a successive procedure including precipitation with bovine serum albumin at pH 4.9, centrifugation to give tannin-BSA pellet, dissolution in TEA buffer containing 5% TEA (v/v) and 10% SDS (w/v), and reaction with ferric chloride reagent (10 mM FeCl 3 in 0.01 N HCl). The product was measured for the absorbance at 510 nm. Catechin was used as the standard. Data were measured in three replications and expressed as mg of catechin equivalent (CE) per liter of wine.

Chromatic Characteristics
For CIE colorimetry, color parameters (L *, a *, and b *) were recorded in a ColorQuest HunterLab spectrocolorimeter equipped with HunterLab Universal Software Version 3.0 (HunterLab, Hunter Associates Laboratories Inc., Reston, VA, USA). Color difference (∆E) between two samples was calculated according to Equation (5): For spectrometry, wine sample was filtered through 0.45 µm membranes and detected for the absorbances at 420, 520, and 700 nm in a 0.2 cm path-length quartz cuvette in a Spectrophotometer U-1800 system (Hitachi, Tokyo, Japan). The color density and hue were calculated as Equations (6) and (7) [5], where A 420 , A 520 , and A 700 were the absorbances at 420, 520, and 700 nm, respectively.

Sensory Evaluation of Red Wine
Consumer acceptance and preference tests were performed on four kinds of wines made of winter (WIN) or summer (SUM) grapes fermented with parent strain (S. cerevisiae BCRC 21685, SC) or mutant CM8. A total of 73 consumer-type panelists (22 males and 51 females) included staffs and students from the Department of Food and Nutrition, Providence University (Taiwan). All tested samples were adjusted to 3 • Brix, according to the result of consumer preference pretest that slightly sweetened wine was more popular than blank wine, and kept in a 4 • C refrigerator before sensory evaluation. Consumer acceptance test was done by 9-point hedonic test with the following standards. The consumer acceptance index was calculated as 100 × means of given points/9. Consumer preference test was done with Rank preference, ISO standards. The evaluated characteristics of wine samples included color, aroma, flavor, acidity, sweetness, astringency, and overall preference.

Statistical Analysis
For chemical compositions, analysis of variance was performed by one-way analysis of variance (one-way ANOVA) procedures using SPSS 9.3 software (SPSS Inc. Chicago, IL, USA). Duncan's new multiple-range test was used to determine the differences among means. A significance level at p < 5% was adopted for all comparisons. XLSTAT software (Addinsoft Inc., New York, NY, USA) was applied for the results of sensory evaluation. ANOVA and Honestly Significant Difference analysis were conducted for consumer acceptance data (9-point hedonic data and acceptance index). Principal component analysis (PCA) was performed on the characteristics of red wines. And, Friedman test for Rank preference, ISO standard was conducted on consumer preference results.

Mannoproteins Contents of S. Cerevisiae Mutants
Colonies, grew in killer-9 toxin containing YPD Figure 1 shows the colonies of parent strain, S. cerevisiae BCRC 21685 (SC), and its mutant strains grown on k9-overlayered agar plate, where mutants M1 and M7 were induced by UV light; and CM3, CM5 and CM8 were by EMS mutagenesis. Generally, five mutants (M1, M7, CM3, CM5, and CM8) grew well at 10 −2 − 10 −4 dilutions (D2-D4) and were apparently more tolerant toward killer-9 toxin than SC, which grew badly at 10 −2 dilution (D2). Accordingly, these mutants were further examined for the monosaccharide compositions and mannoproteins contents in cell wall. Table 1 displays that the cell wall of SC possessed, in average, 151 mg/g glucose (G) and 298 mg/g mannose (M), giving an averaged M/G molar ratio = 1.97. Those of mutants M1, M7, CM3, and CM5 displayed generally 98-126 mg/g G, 215-279 mg/g M, and a M/G molar ratio in the range of 2.06-2.37, not significantly (p > 0.05) different from the SC. Interestingly, mutant CM8 cell wall had in average 111.9 mg/g G, 477.6 mg/g M, leading to a M/G molar ratio of 4.27, significantly (p < 0.05) higher than the SC and the other mutants. That is, mannose in the carbohydrate moiety of mannoproteins accounted for 81% for CM8, in contrast to SC (66%). The mannoprotein content of CM8 cell wall was in average 386.8 mg/g, significantly (p < 0.05) greater than the SC (340 mg/g). Accordingly, CM8 mutant was used as a starter to produce red wine, in comparison with SC.   Table 2 shows the characteristics of four different red wines fermented by parent strain (SC) or mutant (CM8) on winter grapes (WIN) or summer grapes (SUM). Evidently, all wines studied showed similar pH values (3.93-3.94) and alcohol concentrations (in average 117 g/L for SC and 123 g/L for CM8). The titratable acidity was significantly (p < 0.05) greater for SC wines (1.18-1.23 g TAE/L) than for CM8 wines (1.01-1.09 g TAE/L). The wines made from winter grapes (SC-WIN and CM8-WIN) contained significantly (p < 0.05) higher contents of total phenolics (TPC), tartaric esters (TEC), total flavonols (TFC), anthocyanins (indicated by AC M3G and A ace ), and total tannins (TTC), and appeared with significantly (p < 0.05) greater a *, b *, and color density together with lessened L * and color hue, in contrast to their SUM counterparts. Focusing on the WIN wines, both SC-WIN and CM8-WIN exhibited very similar TPC (1409-1439 mg GAE/L), TEC (774-784 mg CAE/L), TFC (255-283 mg QE/L), AC M3G (11.01-11.17 mg M3G/L), A ace (1.436-1.484), and color hue (1.09-1.10). In difference, CM8-WIN showed significantly (p <0.05) greater total anthocyanins (A ace ), TTC, a *, b *, and color density, than those did SC-WIN. For anthocyanins types, the co-pigmented, monomeric, and polymeric ones accounted for 0.8-3.1%, 35.5-37.4%, and 61.0-62.5%, respectively, for all wines. Basically, A ace value was found very closely related to AC M3G : A ace = 0.139×AC M3G -0.0855 (R 2 = 0.995). Besides, the color differences between WIN and SUM wines (∆E WIN-SUM ) appeared to be 28.73 and 29.72 for SC and CM8, respectively, significantly greater than those between SC and CM8 samples (∆E SC-CM8 = 1.64 and 0.63 for WIN and SUM, respectively) (data not tabulated).  Figure 2A shows the results of consumer acceptance test for four wines. It is indicated that, for color, both WIN wines (CM8-WIN and SC-WIN) got generally more panelists for high preference scores, in contrast to those did SUM wines (CM8-SUM and SC-SUM). For aroma ( Figure 2B), all samples showed generally similar preference profiles. For flavor ( Figure 2C), both CM8 wines (CM8-SUM and CM8-WIN) were evaluated with high scores by more panelists than their SC counterparts. And, for overall preference ( Figure 2D), both CM8 wines got more panelists for high scores than did SC wines, in which the sum of panelists scored 7-9 for CM8-SUM, CM8-WIN, SC-SUM, and SC-WIN were 13, 21, 11, and 16, respectively.  Table 2. Table 3 shows the consumer acceptance and rank preference for the wines studied. For color, both WIN wines obtained significantly (p < 0.05) higher preference scores than both SUM wines. For flavor, both CM8 wines got scores significantly greater than both SC wines. For aroma and overall preference, all wines get scores insignificantly different between each other. The resultant acceptance index (AI) was in the order of CM8-WIN (65) > SC-WIN (62) > CM8-SUM (59) > SC-SUM (58). The sum of preference rank by Friedman test was the lowest, i.e., most preferred, for CM8-WIN, followed by SC-WIN, CM8-SUM, and SC-SUM, agreeing with AI. Generally, the wines showed similar degrees of sweetness, acidity, and astringency. Conclusively, CM8-WIN, the topped one rich in total phenolics, tartaric esters, total anthocyanins, and tannins contents, displayed the greatest colorimetric a *, b *, color density, and highest consumer acceptance in color, flavor, aroma, overall, AI, and consumer preference among the wines examined.  Table 2. B Data are presented as means ± S.D. (n = 73); data followed with different superscript letters (a-b) differ significantly (p < 0.05) in the same row. C AI = acceptance index =100 *(averaged scores/9) based on Hedonic 9-point tests for color, flavor, aroma, and overall preference. D By Friedman test, the lower the sum of ranks, the higher was the preference by the consumers. Data in parenthesis are means of ranks (n = 73). Figure 3 depicts the PCA results on the chemical, colorimetric, and some sensory data of the red wines studied. It is evident that factors F1 and F2 explained 92.67% of data variations, where F1 accounted for 75.85%; and F2 did for 16.82%. Active variables, such as degree of sweetness, pH, degree of acidity, various phenolic compositions (total phenolics, tartaric esters, flavonols, anthocyanins, and tannins), and degree of astringency, contributed positively to F1, where degree of astringency did negatively to F2. Active variables L * value, color hue, and titratable acidity contributed negatively to F1, where titratable acidity did negatively to F2. Briefly, F1 concerned mainly color and phenolic compositions (including total phenolics, flavonols, tartaric esters, anthocyanins, and tannins) and was positive for high color (indicated by high color density, a *, and b *) and high phenolic compositions. F2 concerned mainly flavor and titratable acidity and positive for high flavor and low titratable acidity. Quadrant I with CM8-WIN was characterized with high color, high phenolics, and high flavor. Quadrant II with CM8-SUM was with low color, low phenolics, and high flavor. Quadrant III with SC-SUM was with low color, low phenolics, and low flavor. And, quadrant IV with SC-WIN was with high color, high phenolics, and low flavor. Besides, CM8-SUM in quadrant II, in the opposite side of quadrant IV with vector degree of astringency, showed low astringency.

Principal Factors Governing Sensory Preference of Wines
Interestingly, the PCA biplot ( Figure 3) illustrates that the vectors of phenolic compositions (total anthocyanins, flavonols, anthocyanins, total phenolics, tartaric esters, and tannins) and CIE parameters (a *, b *, and color density) were very close, implying that these parameters were closely correlated to each other. Total phenolic content (TPC) could be a key index for describing the variations in the above phenolic compositions and CIE parameters. The results are shown in Table 4 Table 2. Table 4. Changes in various phenolic compositions, colorimetric parameters, and sensory color score of four red wines studies with respect to total phenolics content (TPC).

Variable
Function

Discussion
Evidently, the mannose content in SC mannoproteins was 66% and increased to 81% for that in CM8 mannoproteins. The mannose content in mannoproteins from SC, a wine S. cerevisiae strain, is clearly greater than those from baker's S. cerevisiae strains (39-55%) [25,26]. The mannose content can be increased to 93-97% by isolation assisted with β-(1,3)-glucanase action [25]; however, enzymatic treatments are expansive for industry.
Comparatively, the WIN wines made from Kyoho grape show greater levels of total phenolics, tartaric esters and flavonols, lower titratable acidity, notably lesser anthocyanins level, and comparable tannins content range, in contrast to 78 commercial red wines [5], which exhibit generally pH 3.66-3.72; titratable acidity = 5.80-6. There are some similarities and differences between the PCA biplots for the wines studied and 173 commercial wines from British Columbia in Canada [5]. In similarity, all phenolic compositions examined, CIE a *, b * and color density, and sensory attribute astringency contribute positively to factor F1. While L* and color hue contribute negatively to F1 but positively to F2. Quadrant I is characterized with high phenolics and high tannins; quadrant II, low color (and low phenolics); quadrant III, low phenolics (and low color); and quadrant IV, high color and high anthocyanins. In difference, in this study vector astringency was in quadrant IV, away from vector tannins (quadrant I), pointing to that these two variables are not closely correlated to each other, different from the cases of regular wines that astringency is mostly dependent on tannins level [5,24]. Vectors total phenolics, tartaric esters, flavonols, anthocyanins, and color density, a *, and b * were very close to each other and located in the center between quadrants I and IV, i.e., contributed almost equally to both quadrants concerning CM8-WIN and SC-WIN. Vector titratable acidity was in quadrant III, not in quadrant I as reported for commercial wines [5]. All these differences could be reasonably linked to the effects of CM8 high-M mannoproteins by interactions with wine compositions (tannins, flavonols, anthocyanins, tartaric acid, and aroma), especially tannins, in addition to the effects of grape compositions.
Focusing on the mannoproteins effects, the promising effects of enriched high-M mannoproteins from CM8 could be responsible for the high-quality attributes (color, flavor, and consumer acceptance and preference) of CM8-WIN wines, which possessed greater TTC and slightly higher A ace (total anthocyanins), in contrast to SC-WIN. These effects are similar to those by addition of yeast mannoproteins on reducing astringency, improving sensory characteristics, and increasing aroma persistence for red wines [4,7,27], stabilizing anthocyanins-dominated color, and improving the taste of blueberry wine [10]. Mannoproteins-tannins interactions seem to be noticeable in this study, agreeing with the binding abilities of baker's yeast mannoproteins that are stronger for tannins than the other phenolic compounds [26]. However, the potential interactions of mannoproteins with the other wine compositions, such as proanthocyanins [4,6], flavonols [7], tartaric acid [11], and aroma compounds [2], may not be neglected. These interactions facilitate the solubilization, rather than precipitation, of the related compositions in wine. This is further confirmed by the observation that CM8 wine mashes showed half-volume precipitates of that SC in the end of fermentation (data not shown).
Briefly, using a wine starter of higher-M mannoproteins (>80% mannose), like S. cerevisiae CM8, is more feasible than those of lower-M mannoproteins, such as SC (66% mannose), or baker's yeasts (50-70% mannose) to improve the overall quality attributes of red wines. CM8 mutant may be a good starter competitive and beneficial for wine industry.

Conclusions
This study has discovered a new S. cerevisiae mutant CM8 with overexpressed high-M mannoproteins. Using CM8 or parent S. cerevisiae (SC) as starters for alcoholic fermentation on winter Kyoho grapes (WIN), the resultant wines appeared to be of significantly greater total tannins and total anthocyanins levels, higher color, and higher flavor for CM8-WIN than for SC-WIN wines. By PCA biplots, CM8-WIN wine was classified in quadrant I characterized with high color, high phenolics, and high flavor; and SC-WIN in quadrant IV, high color, high phenolics, low flavor, and high astringency. The improvements in these wine characteristics by CM8, in contrast to SC, could be attributed to its high-M mannoproteins via the interactions with related wine compositions, resembling those by addition of exogenous yeast mannoproteins reported. Conclusively, using CM8 as a starter to ferment Kyoho grapes is feasible to improve wine quality in situ during fermentation and competitive to the effects of adding exogenous mannoproteins in winemaking. More investigations on the effects of CM8 mutant or other yeast strains with various mannose contents in mannoproteins on the quality attributes of diverse wines are needed in the future.

Conflicts of Interest:
The authors have declared that they have no conflict of interest with respect to the work described in this manuscript.