1H NMR-Based Metabolomic Approach for Understanding the

Anal. Chem. 2009, 81, 1137–1145

1

H NMR-Based Metabolomic Approach for Understanding the Fermentation Behaviors of Wine Yeast Strains Hong-Seok Son,† Geum-Sook Hwang,‡ Ki Myong Kim,† Eun-Young Kim,‡ Frans van den Berg,§ Won-Mok Park,† Cherl-Ho Lee,*,† and Young-Shick Hong*,† School of Life Science and Biotechnology, Korea University, 5-1, Anam-dong, Sungbuk-gu, Seoul 136-701, Republic of Korea, Korea Basic Science Institute, Anam-dong, Sungbuk-gu, Seoul 136-713, Republic of Korea, and Department of Food Sciences, Faculty of Life Sciences, University of Copenhagen, Denmark 1

H NMR spectroscopy coupled with multivariate statistical analysis was used for the first time to investigate metabolic changes in musts during alcoholic fermentation and wines during aging. Three Saccharomyces cerevisiae yeast strains (RC-212, KIV-1116, and KUBY501) were also evaluated for their impacts on the metabolic changes in must and wine. Pattern recognition (PR) methods, including PCA, PLS-DA, and OPLS-DA scores plots, showed clear differences for metabolites among musts or wines for each fermentation stage up to 6 months. Metabolites responsible for the differentiation were identified as valine, 2,3-butanediol (2,3-BD), pyruvate, succinate, proline, citrate, glycerol, malate, tartarate, glucose, N-methylnicotinic acid (NMNA), and polyphenol compounds. PCA scores plots showed continuous movements away from days 1 to 8 in all musts for all yeast strains, indicating continuous and active fermentation. During alcoholic fermentation, the highest levels of 2,3-BD, succinate, and glycerol were found in musts with the KIV-1116 strain, which showed the fastest fermentation or highest fermentative activity of the three strains, whereas the KUBY-501 strain showed the slowest fermentative activity. This study highlights the applicability of NMRbased metabolomics for monitoring wine fermentation and evaluating the fermentative characteristics of yeast strains. Wine consists of a number of metabolites that either originate from grapes or are produced during alcoholic fermentation by yeast and malolactic fermentation by lactic acid bacteria. It is wellknown that the “terroir”, which accounts for the factors of climate, soil, and cultural practices, affects the chemical compositions of grapes and that grape varieties also differ in their compositions.1-3 Although grapes’ chemical constituents affect the wine, the vast * To whom correspondence should be addressed. E-mail: [email protected]. † Korea University. ‡ Korea Basic Science Institute. § University of Copenhagen. (1) Pereira, G. E.; Gaudillere, J. P.; Van Leeuwen, C.; Hilbert, G.; Lavialle, O.; Maucourt, M.; Deborde, C.; Moing, A.; Rolin, D. J. Agric. Food Chem. 2005, 53, 6382–6389. 10.1021/ac802305c CCC: $40.75  2009 American Chemical Society Published on Web 12/30/2008

majority of chemicals found in wine are the metabolic byproduct of yeast activity during fermentation. Therefore, it is important to characterize the winemaking abilities of wine yeasts to ensure reproducible fermentation performance and product quality. Preferred wine yeasts have rapid initiation of fermentation, high fermentation efficiency, high ethanol tolerance, a low temperature optimum, high glycerol production, high genetic stability, and low biogenic amine formation.4 The fermentation properties of wine yeasts have been genetically defined. Major changes in gene expressions occur during alcoholic fermentation as yeasts adapt to changes in nutritional, environmental, and physiological conditions, including general stresslike-starvation and ethanol stress.5,6 Global gene expression analyses have been carried out in recent years to characterize the physiological properties of natural and commercial yeasts,5,7-9 as well as genetically and metabolically engineered yeasts.10,11 These analyses have also investigated responses of yeasts to fermentation stress.6,12-14 The effects of carbon source perturbations on gene and protein expressions in the yeast Saccharomyces cerevisiae have been reported, which illustrates the power of integrating genomic and proteomic data for comprehensive characterizations of yeast.15 (2) Pereira, G. E.; Gaudillere, J. P.; Pieri, P.; Hilbert, G.; Maucourt, M.; Deborde, C.; Moing, A.; Rolin, D. J. Agric. Food Chem. 2006, 54, 6765–6775. (3) Pereira, G. E.; Gaudillere, J. P.; Van Leeuwen, C.; Hilbert, G.; Maucourt, M.; Deborde, C.; Moing, A.; Rolin, D. J. Int. Sci. Vigne Vin. 2007, 41, 103– 109. (4) Jackson, R. S. Wine Science; Academic Press: San Diego, CA, 2000. (5) Rossignol, T.; Dulau, L.; Julien, A.; Blondin, B. Yeast 2003, 20, 1369–1385. (6) Marks, V. D.; Sui, S. J. H.; Erasmus, D.; van der Merwe, G. K.; Brumm, J.; Wasserman, W. W.; Bryan, J.; van Vuuren, H. J. J. FEMS Yeast Res. 2008, 8, 35–52. (7) Wu, H.; Zheng, X. H.; Araki, Y.; Sahara, H.; Takagi, H.; Shimoi, H. Appl. Environ. Microbiol. 2006, 72, 7353–7358. (8) Cavalieri, D.; Townsend, J. P.; Hartl, D. L. Proc. Natl. Acad. Sci. U.S.A. 2000, 97, 12369–12374. (9) Varela, C.; Cardenas, J.; Melo, F.; Agosin, E. Yeast 2005, 22, 369–383. (10) Branduardi, P.; Smeraldi, C.; Porro, D. J. Mol. Microb. Biotech. 2008, 15, 31–40. (11) Cebollero, E.; Gonzalez-Ramos, D.; Tabera, L.; Gonzalez, R. Biotechnol. Lett. 2007, 29, 191–200. (12) Pigeau, G. M.; Inglis, D. L. J. Appl. Microbiol. 2007, 103, 1576–1586. (13) Beltran, G.; Novo, M.; Leberre, V.; Sokol, S.; Labourdette, D.; Guillamon, J. M.; Mas, A.; Francois, J.; Rozes, N. FEMS Yeast Res. 2006, 6, 1167– 1183. (14) Zuzuarregui, A.; Monteoliva, L.; Gil, C.; del Olmo, M. L. Appl. Environ. Microbiol. 2006, 72, 836–847.

Analytical Chemistry, Vol. 81, No. 3, February 1, 2009

1137

Metabolomics or metabonomics in the -omics era is a promising new approach aimed at improving our understanding of metabolic perturbations in drug toxicity,16-18 disease status,19 dietary intervention,20-22 and plant primary and secondary metabolism.23,24 Most recently, human metabolic phenotype diversity, based on 1H NMR spectroscopy, was reported for 4 630 participants’ urinary excretion of alanine and hippurate along with their associations to diet, gut microbial activities, and individual blood pressure.25 The advent of powerful chemical analytical tools, such as mass spectrometry (MS) and nuclear magnetic resonance (NMR), and the development of techniques for multivariate statistical modeling have led to increased adoption of large-scale metabolic analyses.26 Multivariate statistical pattern recognition (PR) methods, such as principal components analysis (PCA), partial least-squares discriminant analysis (PLS-DA), and orthogonal projections to latent structures discriminant analysis (OPLS-DA), reduce the dimensionality of complex data sets and thereby facilitate the visualization of inherent patterns in the data. Pereira et al.1,2,27 reported on the effects of microclimate, vintage, and soil for metabolites of grape berries using 1H NMR spectroscopy and PR methods. 1H NMR spectroscopy was also used to monitor the time-course for the evolution of malic and lactic acids during alcoholic and malolactic fermentations of grape must.28 However, the evolutions of these acids were quantitative and targeted but were not statistical or global. In our previous study for the characterization of wines by 1H NMR global metabolite profiling, we showed metabolic differences in wines from different grape varieties and production (15) Griffin, T. J.; Gygi, S. P.; Ideker, T.; Rist, B.; Eng, J.; Hood, L.; Aebersold, R. Mol. Cell. Proteomics 2002, 1, 323–333. (16) Clayton, T. A.; Lindon, J. C.; Cloarec, O.; Antti, H.; Charuel, C.; Hanton, G.; Provost, J. P.; Le Net, J. L.; Baker, D.; Walley, R. J.; Everett, J. R.; Nicholson, J. K. Nature 2006, 440, 1073–1077. (17) Coen, M.; Hong, Y. S.; Clayton, T. A.; Rohde, C. M.; Pearce, J. T.; Reily, M. D.; Robertson, D. G.; Holmes, E.; Lindon, J. C.; Nicholson, J. K. J. Proteome Res. 2007, 6, 2711–2719. (18) Nicholson, J. K.; Connelly, J.; Lindon, J. C.; Holmes, E. Nat. Rev. Drug Discovery 2002, 1, 153–161. (19) Brindle, J. T.; Antti, H.; Holmes, E.; Tranter, G.; Nicholson, J. K.; Bethell, H. W. L.; Clarke, S.; Schofield, P. M.; McKilligin, E.; Mosedale, D. E.; Grainger, D. J. Nat. Med. 2002, 8, 1439–1444. (20) Wang, Y. L.; Tang, H. R.; Nicholson, J. K.; Hylands, P. J.; Sampson, J.; Holmes, E. J. Agric. Food Chem. 2005, 53, 191–196. (21) Solanky, K. S.; Bailey, N. J.; Beckwith-Hall, B. M.; Bingham, S.; Davis, A.; Holmes, E.; Nicholson, J. K.; Cassidy, A. J. Nutr. Biochem. 2005, 16, 236– 244. (22) Solanky, K. S.; Bailey, N. J. C.; Beckwith-Hall, B. M.; Davis, A.; Bingham, S.; Holmes, E.; Nicholson, J. K.; Cassidy, A. Anal. Biochem. 2003, 323, 197–204. (23) Hagel, J. M.; Weljie, A. M.; Vogel, H. J.; Facchini, P. J. Plant Physiol. 2008, 147, 1805–1821. (24) Zulak, K. G.; Weljie, A. M.; Vogel, H. J.; Facchini, P. J. BMC Plant Biol. 2008, 8, 5. (25) Holmes, E.; Loo, R. L.; Stamler, J.; Bictash, M.; Yap, I. K. S.; Chan, Q.; Ebbels, T.; De Iorio, M.; Brown, I. J.; Veselkov, K. A.; Daviglus, M. L.; Kesteloot, H.; Ueshima, H.; Zhao, L. C.; Nicholson, J. K.; Elliott, P. Nature 2008, 453, 396–U50. (26) Crockford, D. J.; Holmes, E.; Lindon, J. C.; Plumb, R. S.; Zirah, S.; Bruce, S. J.; Rainville, P.; Stumpf, C. L.; Nicholson, J. K. Anal. Chem. 2006, 78, 363–371. (27) Pereira, G. E.; Gaudillere, J. P.; van Leeuwen, C.; Hilbert, G.; Maucourt, M.; Deborde, C.; Moing, A.; Rolin, D. Anal. Chim. Acta 2006, 563, 346– 352. (28) Avenoza, A.; Busto, J. H.; Canal, N.; Peregrina, J. M. J. Agric. Food Chem. 2006, 54, 4715–4720.

1138


areas.29 We report here a statistical evolution of metabolites in grape must during alcoholic fermentation and aging and the evaluation of the fermentative performances of yeast strains using 1H NMR-based metabolomics coupled with multivariate statistical analyses. MATERIALS AND METHODS Yeast Strains. Two commercial Saccharomyces cerevisiae yeast strains (Lalvin RC-212 and K1V-1116) and one S. cerevisiae strain (KUBY-501) isolated from Korean wild grape, Raspberry (Rubus corenus) (W. M. Park et al., unpublished data, School of Life Science and Biotechnology, Korea University), were used for vinification. Briefly, the S. cerevisiae KUBY-501 strain has a colony with convex morphology and white color and can survive at 80 °C. KUBY-501 gives raspberry wine the body, slight sweetness, and strong flavor of the raspberry grape. On the basis of genetic analysis including the sequences analyses of 18S rRNA and the internal transcribed spacer (ITS) region,30,31 it has been determined to be a S. cerevisiae new species and thus named S. cerevisiae KUBY-501. Vinification. Muscat Bailey A grapes (Vitis labruscana) were harvested in 2007 (100 kg) and crushed. The must was distributed into 15 25 L plastic tanks, producing 5 batches for each of the 3 yeast strains. Dried yeasts were first multiplied on YPD medium (1% w/v yeast extract, 1% w/v peptone, and 2% w/v dextrose) solidified with 2% (w/v) agar. The yeast was activated in YPD medium and then cultured in a 1:1 mixture of YPD and grape juice at 25 °C for 24 h to obtain a final cell count of 2 × 106 cells/ mL. Potassium metabisulfite (100 mg/kg grape) was added to the must, and starters for the 3 yeast strains were inoculated at 2 × 106 cells/mL. Alcoholic fermentation with the 3 yeast strains was carried out in the carboys at 20-22 °C for 8 days. After completion of alcoholic fermentation, the musts were transferred into 15 4 L glass carboys at days 9 and then racked at 3 and 6 months, followed by sampling the must or wine. 1 H NMR Spectroscopic Analysis of Musts and Wines. A volume of 1 mL of must or wine was lyophilized in a 1 mL Eppendorf tube and dissolved in 99.9% deuterium oxide (400 µL, D2O), mixed with 400 mM oxalate buffer (140 µL, pH 4.0), and 5 mM sodium 2,2-dimethyl-2-silapentane-5-sulfonate (60 µL, DSS, 97%) and then centrifuged at 13 000 rpm for 10 min. Supernatants (550 µL) were transferred into 5 mm NMR tubes. D2O and DSS provided a field frequency lock and chemical shift reference (1H, δ 0.00), respectively. 1H NMR spectra were acquired on a VnmrS-600 MHz NMR spectrometer (Varian Inc., Palo Alto, CA) operating at 599.84 MHz 1H frequency and a temperature of 298 K, using a triple resonance 5-mm HCN salt tolerant cold probe. A NOESYPREAST pulse sequence was applied to suppress the residual water signal. For each sample, 16 transients were collected into 32 K data points using a spectral width of 9615.4 Hz with a relaxation delay of 1.5 s, an acquisition time of 4.00 s, and a mixing time of 400 ms. A 0.3 Hz line-broadening function was applied to all spectra prior to Fourier transformation (FT). (29) Son, H. S.; Kim, K. M.; van den Berg, F.; Hwang, G. S.; Park, W. M.; Lee, C. H.; Hong, Y. S. J. Agric. Food Chem. 2008, 56, 8007–8016. (30) Mankin, A. S.; Skryabin, K. G.; Rubtsov, P. M. Gene 1986, 44, 143–145. (31) Naumov, G. I.; James, S. A.; Naumova, E. S.; Louis, E. J.; Roberts, I. N. Int. J. Syst. Evol. Microbiol. 2000, 50, 1931–1942.

NMR Data Reduction and Preprocessing. All NMR spectra were phased and baseline corrected by Chenomx NMR suite4.6 software, professional edition (Chenomx Inc., Canada). The NMR spectral data was reduced into 0.001 ppm spectral buckets, while the region corresponding to water (4.6-4.8 ppm) was removed. In addition, the regions for residual ethanol (1.15-1.20 and 3.59-3.72 ppm) from incomplete removal during lyophilization and for DSS (-0.5-0.7 ppm) were also removed. The spectra were then normalized to the total spectral area and converted to ASCII format. The ASCII format files were imported into MATLAB (R2006a, Mathworks, Inc., 2006), and all spectra were aligned using the correlation optimized warping (COW) method.29,32 The resulting data sets were then imported into SIMCA-P version 12.0 (Umetrics, Umeå, Sweden) for multivariate statistical analysis. Signal assignment for representative samples was facilitated via acquisition of two-dimensional (2D) total correlation spectroscopy (TOCSY), heteronuclear multiple bond correlation (HMBC), heteronuclear single quantum correlation (HSQC), spiking experiments, and comparisons to literature. In addition, Chenomx NMR suite4.6 software was utilized to assign the metabolites in wine. Multivariate Data Analysis. The mean center was applied for all multivariate analysis by SIMCA-P version 12.0 (Umetrics, Sweden). Principal components analysis (PCA), an unsupervised pattern recognition method, was performed to examine the intrinsic variation in the data set. To maximize the separation between samples, partial least-squares discriminant analysis (PLSDA), was applied. The PLS-DA is the regression extension of PCA, which gives the maximum covariance between the measured data (X variable, metabolites in NMR spectra) and the response variable (Y variable, NMR spectral intensities). Orthogonal projections to the latent structures discriminant analysis (OPLS-DA) model was also applied to remove noncorrelated variation in X variables to Y variables or for variability in X that is orthogonal to Y. Hotelling’s T2 region, shown as an ellipse in the scores plots, defines the 95% confidence interval of the modeled variation.33 The quality of the models was described by R2 and Q2 values. R2 is defined as the proportion of variance in the data explained by the models and indicates goodness of fit, and Q2 is defined as the proportion of variance in the data predictable by the model and indicates predictability.25,34 Chemicals. All chemical reagents were of analytical grade. D2O (99.9%), DSS (97%), 2-phenylethanol, and gallic acid were purchased from Sigma (St. Louis, MO). RESULTS Figure 1 shows representative 600 MHz 1H NMR spectra of musts up to day 8 of alcoholic fermentation using strain RC212. A wide range of metabolites can be assigned in each spectral matrix that provides complementary information for global changes of metabolites during alcoholic fermentation. Visual inspection of the NMR spectra revealed marked changes of several metabolites. These included increased levels of 2,3butanediol (2,3-BD), lactate, unknown compound (U), pyruvate, succinate, malate, citrate, glycerol, tartarate, polyphenols, and N-methylnicotinic acid (NMNA) and decreased levels of (32) Larsen, F. H.; van den Berg, F.; Engelsen, S. B. J. Chemometrics 2006, 20, 198–208. (33) Hotelling, H. Ann. Math. Stat. 1931, 2, 360–378. (34) Trygg, J.; Wold, S. J. Chemom. 2002, 16, 119–128.

glucose during alcoholic fermentation. To provide comparative interpretations for these metabolic changes at each fermentation stage, a series of pattern recognition methods were employed. As described in Materials and Methods, PCA, PLSDA, or OPLS-DA35 were applied to NMR spectral data to visualize metabolic differentiation of musts or wines obtained at each fermentation stage and during the aging period. The assignments of metabolites were verified by two-dimensional NMR experiments as shown in Figure S-1 in the Supporting Information. During alcoholic fermentation, marked increases in the levels of oligosaccharides (O) that were tentatively assigned and unknown compounds (U1) were observed. Metabolic Changes in Musts During Alcoholic Fermentation. A PCA scores plot (Figure 2) shows clear day-to-day differences between musts during alcoholic fermentation using strain RC-212. Movements of the plots from the left to the right indicate continuous metabolic changes or fermentation. Figure 3A-F shows PCA scores plots that differentiate musts before and after 24 h of alcoholic fermentation and reveal their day-to-day pairwise comparisons (days 2-8). Two samples, one each from days 7 and 8, were excluded because they clustered with 2 different musts at days 7 and 8, indicating one each for slow and fast fermentation. After the exclusion, the PCA was regenerated (Figure 3F). For day-to-day comparisons from the PCA model, the statistics for differentiating musts during fermentation showed high goodness of fit and predictability with R2 values from 0.77 to 0.96 and Q2 values from 0.69 to 0.92. However, comparison of musts at days 2 and 3 only gave an R2 value of 0.66 and a Q2 value of 0.38. These lower R2 and Q2 values for days 2 and 3 might indicate that metabolic changes at later fermentation times were larger than those during early fermentation. To identify the metabolites responsible for the differentiation in PCA scores plots over alcoholic fermentation stages, PCA loadings plots were generated (Figure 3G-L). The upper sections of the loadings plots represent metabolites that were higher in musts after alcoholic fermentation for 24 h, whereas the lower sections reveal metabolites that were lower. The differentiations of musts before and after 24 h were due to increases of U1, valine, 2,3-butanediol (2,3-BD), lactate, pyruvate, proline, succinate, glycerol, malate, citrate, tartarate, unidentified compound (5.18 ppm), N-methylnicotinic acid (NMNA), and polyphenols and decreases of R- and β-glucose, γ-aminobutyrate (GABA), and alanine (Figure 3G-L). Dramatic consumption of glucose and the production of glycerol, 2,3-BD, lactate, and succinate by yeast were observed in the statistical loadings plots over the entire alcoholic fermentation period, showing an active, ongoing fermentation. Metabolic Changes in Wines During Aging. We analyzed the metabolic changes in wines during aging that were vinified with strain RC-212. PCA was initially applied to the entire metabolomic data obtained from musts at the end of alcoholic fermentation and from wines at 3 and 6 months of aging. A Q2 value was calculated for the PCA model to assess any significance of differentiation. However, no significant differentiations were found between must at day 8 and wines at 3 and 6 months, using the PCA model, which reflects a (35) Jackson, J. E. A. A Users Guide to Principal Components; Wiley: New York, 1991.


1139

Figure 1. tk;4Representative 600 MHz 1H NMR spectra of musts obtained at days 2 (A), 3 (B), 5 (C), and 7 (D) during alcoholic fermentation. / denote residual ethanol peaks that remained during lyophilization, which were excluded in multivariate statistical analysis.

poor predictive capability as indicated by a Q2 value of 0.09 (Figure S-2A in the Supporting Information). Partial leastsquares discriminant analysis (PLS-DA) was then used for the differentiation of NMR data from the must and wines. Clear differentiations were observed between must at day 8 and wines at 3 and 6 months in the PLS-DA score plot (Figure S-2B in the Supporting Information). Compared to the PCA model, this PLS-DA model showed improved predictability with a Q2 value of 0.54. An OPLS-DA scores plot is shown in Figure S-2C in the Supporting Information. This model gave a more discernible differentiation between the must and wines compared to the PLS-DA scores plot given in Figure S-2B in the Supporting Information. The OPLS-DA model had a higher RX2 value (0.57) and RY2 value (0.93) compared, respectively, to 0.49 and 0.89 for the PLS-DA model. This resulted from the effective 1140


removal of noncorrelated variation in X variables (metabolites in NMR spectra) to Y variables (NMR spectral intensities) or of variability in X that is orthogonal to Y. The statistics for PCA, PLS-DA, and OPLS-DA are summarized in Table 1. For pair wise comparisons between must at day 8 and wine at 3 months, RX2 increased from 0.47 in the PLSDA model to 0.85 in the OPLS-DA model, RY2 increased from 0.95 to 0.99, and Q2 increased from 0.73 to 0.76. In particular, for wines at 3 and 6 months RX2 increased from 0.39 in the PLS-DA model to 0.52 in the OPLS-DA model and Q2 from 0.37 to 0.84. To provide the pairwise comparisons of wines during aging, an OPLS-DA scores plot was generated, which showed a clear differentiation with high values for the statistical values (Figure S-2C in the Supporting Information). As shown in Figure S-3 in the Supporting Information, OPLSDA scores plots showed clear differentiations between must at

Figure 2. PCA score plot derived from the 1H NMR must and wine spectra for each fermentation day. The movements of the scores plots from left to right indicate continuous metabolic changes and fermentation.

day 8 and wine at 3 months (Figure S-3A in the Supporting Information) and between wines at 3 and 6 months (Figure S-3B in the Supporting Information). An OPLS-DA loadings plot for a model differentiating the 3 month wine from the 8 day must showed that the metabolites responsible for the differentiation were increased levels of U1, leucine, 2,3-BD, proline, succinate, glycerol, polyphenols, and NMNA, together with decreased levels of pyruvate and an unidentified compound (5.18 ppm) (Figure S-3C in the Supporting Information). In addition, the differentiation between wines at 3 and 6 months was caused by decreases of 2,3-BD, tartarate, and proline, together with increases of unidentified compounds that were found at 1.68, 3.00, and 3.48 ppm in wines at 6 months compared to wines at 3 months (Figure S-3D in the Supporting Information). The decreased levels of tartarate in wines at 6 months compared to those in wines at 3 months might be due to cold precipitation of tartarate, as after racking of wine at 3 months it was aged during the winter season in Korea (end of December to March), with daily average temperatures below 10 °C in the storage room. Metabolic Changes by Yeast Strain. Figure S-4 in the Supporting Information shows a PCA scores plot derived for all musts and wines fermented with S. cerevisiae strains RC-212, K1V116, and KUBY-501. As alcoholic fermentation days increased, the PCA scores plots for all musts moved to the right indicating continuous fermentation. It is interesting to note that different fermentation behaviors for each yeast strain are observed by the scores plots. The K1V-1116 strain showed the fastest fermentation or highest fermentative activity, whereas the KUBY-501 strain showed the slowest fermentation. This result demonstrates both fast and slow changes in metabolites during alcoholic fermentation. To investigate metabolic changes in wines for each yeast strain, PCA for musts at day 6 (Figure 4) and OPLS-DA models for wines at 6 months (Figure S-5 in the Supporting Information) were applied. PCA scores plots show clear separations between musts for each yeast strain at day 6 (Figure 4A-C). On PCA loadings plots, the metabolites responsible for the separations of musts with the K1V-1116 strain from musts with the RC-212 (Figure 4D) and KUBY-501 (Figure 4F) strains at day 6 were increased levels of U1, 2,3-BD, proline, alanine, succinate, malate, citrate, glycerol, unidentified compound (5.18 ppm), NMNA, and

polyphenols, together with decreased levels of glucose and pyruvate. In addition, as shown in the PCA loading plot in Figure 4E, U1, 2,3-BD, proline, succinate, malate, citrate, glycerol, unidentified compound (5.18 ppm), NMNA, and polyphenols were increased in must with strain RC-212 compared to those with strain KUBY-501. To compare metabolite changes in wines at 6 months according to yeast strains, OPLS-DA scores plots (parts A and B of Figures S-5 in the Supporting Information) also showed clear separations. However, fewer metabolites contributed to these separations compared to the differences between musts or wines at early fermentation or at 3 months in complementary OPLS-DA loadings plots (Figure S-5D-F in the Supporting Information). Levels of an unknown compound (U1) were still highest in wines from KIV-1116 strain at 6 months. Levels of 2,3-BD, lactate, pyruvate, and polyphenols were highest in wines with strain KUBY-501 at 6 months, while levels of unidentified oligosaccharides (O) were lowest. These results were also observed in wines at 3 months (data not shown). DISCUSSION AND CONCLUSIONS Carbohydrates. The main carbohydrates in grapes are glucose, fructose, rhamnose, arabinose, xylose, sucrose, and pectin. During alcoholic fermentation, consumption of glucose by yeast is slightly faster than fructose.36 As expected, rapid consumption of glucose by yeasts was observed during alcoholic fermentation in this study. The KIV-1116 strain showed the fastest consumption of glucose, while the KUBY-501 strain showed the slowest consumption (Figure 4). In general, polysaccharides, such as pectins and gums, are stable in must and wines. They are less soluble in alcohol, which results in precipitation of up to 50%-80% of their initial concentrations.36 According to recent reports,37,38 soluble polysaccharides in wine are essentially composed of grape cell wall polysaccharides, such as arabinogalactans, arabinogalactan-proteins, and rhamnogalacturonans-II. During alcoholic fermentation, unidentified compounds, including those with typical resonances at 5.18 ppm (doublet), were assigned to oligosaccharides (O) in this study. These oligosaccharides were increased but not consumed by the three yeast strains in all musts (Figure 3). However, decreased levels of these oligosaccharides were observed after 3 months, indicating their slower consumption by yeasts than was typical for monosaccharides, such as glucose (Figure S-3 in the Supporting Information). These oligosaccharides’ levels were lowest in wines obtained at 3 months (data not shown) and at 6 months (Figure S-5 in the Supporting Information) with the KUBY-501 strain. The mean spectra of wines at 6 months might confirm the largest consumption of oligosaccharides by the KUBY-501 strain (Figure S-6 in the Supporting Information). This suggests that KUBY-501 could use oligosaccharides more effectively and thus give wines the most dryness to tastes. Acids. During alcoholic fermentation, yeasts metabolize or synthesize malate.39-41 However, in this study, the skin and pulp were not removed from must during alcoholic fermentation. The synthesis of malate by yeasts via the fumarate or oxaloacetate (36) Margalit, Y. Concepts in Wine Chemistry; The Wine Appreciation Guild: South San Francisco, CA, 2004. (37) Ayestaran, B.; Guadalupe, Z.; Leon, D. Anal. Chim. Acta 2004, 513, 29– 39. (38) Guadalupe, Z.; Ayestaran, B. J. Agric. Food Chem. 2007, 55, 10720–10728.


1141

Figure 3. PCA scores (A-F) and loadings (G-L) plots derived from the 1H NMR must spectra demonstrating significant time-course statistical changes of metabolites during alcoholic fermentation with the RC-212 strain. Upper sections of the loadings plots represent elevated levels of metabolites, whereas the lower sections reveal decreased levels of metabolites on each fermentation day. The statistics for a model differentiating musts during alcoholic fermentation revealed high R2 values, from 0.77 to 0.96, and Q2 values, from 0.69 to 0.92, in PCA models, except for must between days 2 and 3 with a low R2 value of 0.66 and Q2 value of 0.38. The R2 value represents the goodness of fit of the models. The Q2 value represents the predictability of the models. 2,3-BD, 2,3-butanediol; GABA, γ-aminobutyrate.

pathways appeared to be not much more than that released into must by alcoholic extraction without removing the grapes. Levels of tartarate and malate, which are the key acids in grapes or wines, 1142


increased considerably over the entire alcoholic fermentation. This indicated that these acids were mainly released into musts from the grapes during alcoholic extraction rather than from yeasts

Figure 4. PCA scores and loadings plots derived from the 1H NMR spectra of musts at day 6 as pair wise comparisons of musts with yeast strains: (A) PCA score plot showing separation between musts with RC-212 strain (9) and KIV-1116 strain (red b) and (D) complementary loading plot of the first component (PC1); (B) PCA score plot showing separation between musts with RC-212 strain (9) and KUBY-501 strain (blue [) and (E) complementary loading plot of the PC1; (C) PCA score plot showing separation between musts with KIV-1116 strain (red b) and with KUBY-501 strain (blue [) and (F) complementary loading plot of the PC1. PCA models gave high values of R2 from 0.82 to 0.96 and Q2 of 0.69 to 0.93. NMNA, N-methylnicotinic acid; 2,3-BD, 2,3-butanediol. Table 1. PCA, PLS-DA, and OPLS-DA Modeling Results Comparing Must at Day 8 and Wines at 3 and 6 Months with the RC-212 Straina models

PCA or PLS component

PCA PLS-DA OPLS-DA

3 3 2

orthogonal component

RX2

RY2

Q2

2

0.62 0.49 0.57

0.89 0.93

0.09 0.54 0.54

a All values of RX2, RY2, and Q2 are from cumulative values up to the specified component.

metabolism. Their levels did not significantly change after day 8 of fermentation and up to 6 months of aging (Figure S-3 in the Supporting Information) and were not different among the wines aged with the 3 yeast strains for 6 months (Figure S-5 in the Supporting Information). It is likely that the decreased content of tartarate in wines at 3 months compared to wines at 6 months was due to potassium bitartarate and calcium tartarate formation, followed by their precipitation.35,42 This was probably due to cold (39) Rankine, B. C. J. Sci. Food Agric. 1966, 17, 312–316. (40) Fatichenti, F.; Farris, G. A.; Deiana, P.; Ceccarelli, S. Appl. Environ. Microbiol. 1984, 19, 427–429. (41) Schwartz, H.; Radler, F. Appl. Environ. Microbiol. 1988, 27, 553–560. (42) McKinnon, A. J.; Scollary, G. R.; Solomon, D. H.; Williams, P. J. Am. J. Enol. Vitic. 1995, 46, 509–517.

temperatures in the winemaking or storage place for periods between 3 and 6 months in this study. Citrate also primarily originates from grapes. Citrate can be fermented into lactate from sugar by yeast during alcoholic fermentation and by lactic acid bacteria (LAB) during malolactic acid fermentation (MLF).28 Citrate levels continued to increase during alcoholic fermentation, but no significant changes were observed after alcoholic fermentation and up to 6 months of aging. The lack of changes in malate and citrate levels after completion of alcoholic fermentation indicates that there was no bacterial contamination or spontaneous MLF, as LAB completely converts malate or citrate into lactate or other materials during MLF.28,43 As with tartarate and malate, the increase in citrate during alcoholic fermentation was most likely due to extraction of citrate from grapes. It is also likely that the increased levels of lactate were from metabolism of sugars during alcoholic fermentation. γ-Aminobutyrate (GABA) is a four-carbon, nonprotein amino acid conserved from bacteria to plants. The pathway for GABA metabolism is comprised of glutamate decarboxylase, GABA transaminase, and succinic semialdehyde dehydrogenase in plants.44,45 S. cerevisiae utilizes GABA by transaminating a γ-amino (43) Lonvaud-Funel, A. Antonie van Leeuwenhoek 1999, 76, 317–331. (44) Bouche, N.; Fromm, H. Trends Plant Sci. 2004, 9, 110–115. (45) Shelp, B. J.; Bown, A. W.; McLean, M. D. Trends Plant Sci. 1999, 4, 446– 452.


1143

Figure 5. Schematic of metabolic pathways of wine yeast characterized in this study. Arrows indicate increases (v) or decreases (V) of corresponding metabolites during alcoholic fermentation, which were observed by NMR spectroscopy.

group to R-ketoglutarate and degrading the resulting succinate semialdehyde to succinate, which requires GABA transaminase and succinic semialdehyde dehydrogenase, respectively.46 Thus, rapid consumption of GABA by wine yeast by day 4 of alcoholic fermentation indicates high activities of these two enzymes (Figure 3). In addition to GABA, alanine and valine are also used as nitrogen sources during early alcoholic fermentation. However, valine levels were increased after day 4 of alcoholic fermentation. This may indicate that yeast growth reached a plateau at day 5. 2,3-Butanediol and Glycerol. 2,3-Butanediol is a byproduct of alcoholic fermentation that results from the reduction of acetoin. This can derive from the decarboxylation of R-acetolactate, the condensation of active acetaldehyde with acetyl coenzyme, and the condensation of the active acetaldehyde (acetaldehyde-TPP complex) with free acetaldehyde formed from pyruvate.47 Straindependent production of these compounds has been reported for wine yeasts during wine fermentation.48,49 Romano50 also reported that actively fermenting yeasts produced high amounts of 2,3butanediol, whereas weakly fermenting yeasts produced low levels. The production of 2,3-butanediol has, therefore, been used to Pietruszko, R.; Fowden, L. Ann. Bot. 1961, 25, 491–511. Romano, P.; Suzzi, G. Appl. Environ. Microbiol. 1996, 62, 309–315. Romano, P.; Suzzi, G. FEMS Lett. 1993, 108, 23–26. Romano, P.; Suzzi, G.; Brandolini, V.; Menziani, E.; Domizio, P. Lett. Appl. Microbiol. 1996, 22, 299–302. (50) Romano, P.; Granchi, L.; Caruso, M.; Borra, G.; Palla, G.; Fiore, C.; Ganucci, D.; Caligiani, A.; Brandolini, V. Int. J. Food Microbiol. 2003, 86, 163–168. (46) (47) (48) (49)

1144


characterize wine yeasts.51 The highest levels of 2,3-butanediol in musts fermented with the K1V-1116 strain at day 5 (Figure 4) was consistent with the highest fermentative activity of this strain (see PCA score plot of Figure S-4 in the Supporting Information). In contrast, the KUBY-501 strain showed the slowest fermentative activity, resulting in the lowest amounts of 2,3-butanediol. This is consistent with the slowest consumption of glucose by the KUBY501 strain (Figure S-4 in the Supporting Information). However, levels of 2,3-butanediol during aging were highest in wines fermented with the KUBY-501 strain at 3 months (data not shown) and at 6 months (Figure S-5 in the Supporting Information) even though KUBY-501 showed the lowest fermentative activity during alcoholic fermentation (Figure 4). Glycerol levels depend on sulfite concentration, pH, fermentation temperature, aeration, yeast strains, grape variety and ripeness, and nitrogen composition of wine.52-55 These continued to increase up to 3 months in musts or wines with the RC-212 strain. However, there were no variations in glycerol levels between wines at 3 and 6 months with the RC-212 strain. In addition, it was clear that glycerol production depended on the yeast strain, with the highest level of glycerol in musts by KIV1116 at day 5 of alcoholic fermentation. However, the glycerol level did not depend on the yeast strain in wines at 6 months. Significant variations in acetoin and 2,3-butanediol production between yeast strains and their relations to the levels of glycerol produced have been reported.56 Overproduction of glycerol caused an increase in 2,3-butanediol level. These relationships were observed both in wines during aging and musts during alcoholic fermentation in this study. The variations in 2,3-butanediol levels by yeast strains may be a marker for characterizing wine yeasts. Pyruvate, succinate, glycerol, lactate, and 2,3-butanediol levels increased over the entire alcoholic fermentation period, indicating continuously active fermentation. However, these levels did not change during aging. Stability of succinate levels has been reported during aging of wines.57 This suggests that succinate, glycerol, lactate and 2,3-butanediol might be used to characterize fermentative behavior of yeast strains during early wine fermentation. Continuous increases of succinate, U1, lactate, and polyphenols, together with decreased levels of oligosaccharides in wines at 3 months, also suggested that alcoholic fermentation was still occurring after 8 days of alcoholic fermentation. This occurred even after racking off lees or residual skins at day 8, as these compounds were not significantly changed at 6 months. In conclusion, it was clear that of the three yeast strains, the KIV-1116 showed the fastest fermentative behavior with the highest production of glycerol, succinate, lactate, and 2,3-butanediol and the fastest consumption of glucose during alcoholic fermentation. However, the highest production of 2,3-butanediol, lactate, and pyruvate, together with the largest consumption of oligosaccharides, was found in wines with the KUBY-501 strain (51) Romano, P.; Brandolini, V.; Ansaloni, C.; Menziani, E. World J. Microbiol. Biotechnol. 1998, 14, 649–653. (52) Gardner, N.; Rodrigue, N.; Champagne, C. P. Appl. Environ. Microbiol. 1993, 59, 2022–2028. (53) Radler, F.; Shulz, H. Am. J. Enol. Vitic. 1982, 33, 36–40. (54) Rankine, B. C.; Bridson, D. A. Am. J. Enol. Vitic. 1971, 22, 2–12. (55) Albers, E.; Larsson, C.; Liden, G.; Niklasson, C.; Gustafsson, L. Appl. Environ. Microbiol. 1996, 62, 3187–3195. (56) Remize, F.; Roustan, J. L.; Sablayrolles, J. M.; Barre, P.; Dequin, S. Appl. Environ. Microbiol. 1999, 65, 143–149. (57) Thoukis, G.; Ueda, M.; Wright, D. Am. J. Enol. Vitic. 1965, 16, 1–8.

at 6 months (Figures S-5 and S-6 in the Supporting Information). These results indicate slower but more effective fermentation performance of the KUBY-501 strain. In addition, the greatest production of an unknown compound (U1) in must and wine by the KIV-1116 strain could be one characteristic of this strain. These results demonstrate that global analysis of metabolites can provide insights into wine fermentation and the fermentative behaviors of yeast strains through their metabolic pathways (Figure 5). Most recently, the dynamic biochemical compositions within living systems provides a fundamental systems biology, which attempts to synergistically integrate the data sets from gene expression (transcriptomics), protein translation (proteomics), and metabolite network (metabolomics or metabonomics) to provide a more holistic overview of living systems,58-60 such as with Cabernet Sauvignon grape berry.61 However, global systems biology based on integrating data sets have not yet been reported for yeast or (58) Dumas, M. E.; Wilder, S. P.; Bihoreau, M. T.; Barton, R. H.; Fearnside, J. F.; Argoud, K.; D’Amato, L.; Wallis, R. H.; Blancher, C.; Keun, H. C.; Baunsgaard, D.; Scott, J.; Sidelmann, U. G.; Nicholson, J. K.; Gauguier, D. Nat. Genet. 2007, 39, 666–672. (59) Rantalainen, M.; Cloarec, O.; Beckonert, O.; Wilson, I. D.; Jackson, D.; Tonge, R.; Rowlinson, R.; Rayner, S.; Nickson, J.; Wilkinson, R. W.; Mills, J. D.; Trygg, J.; Nicholson, J. K.; Holmes, E. J. Proteome Res. 2006, 5, 2642–2655. (60) Li, M.; Wang, B. H.; Zhang, M. H.; Rantalainen, M.; Wang, S. Y.; Zhou, H. K.; Zhang, Y.; Shen, J.; Pang, X. Y.; Zhang, M. L.; Wei, H.; Chen, Y.; Lu, H. F.; Zuo, J.; Su, M. M.; Qiu, Y. P.; Jia, W.; Xiao, C. N.; Smith, L. M.; Yang, S. L.; Holmes, E.; Tang, H. R.; Zhao, G. P.; Nicholson, J. K.; Li, L. J.; Zhao, L. P. Proc. Natl. Acad. Sci. U.S.A. 2008, 105, 2117–2122. (61) Deluc, L. G.; Grimplet, J.; Wheatley, M. D.; Tillett, R. L.; Quilici, D. R.; Osborne, C.; Schooley, D. A.; Schlauch, K. A.; Cushman, J. C.; Cramer, G. R. BMC Genomics 2007, 8, 429.

alcoholic fermentation in wine research, even though studies for understanding of gene networks or regulation, monitoring of yeast gene expression,5,6,62-64 and proteomic approaches to identify Champagne wine proteins65 are arising. Therefore, metabolomics by global analysis of metabolites highlights a possibility of integrating “omics” to understand metabolic, proteomic, and gene networks in yeast. ACKNOWLEDGMENT G.-S. Hwang contributed equally to this work.The authors acknowledge the Korea University Grant for the research professorship of Dr. Y.-S. Hong, and Dr. G.-S. Hwang is acknowledged for financial support from the KBSI Grant. SUPPORTING INFORMATION AVAILABLE Additional information as noted in text. This material is available free of charge via the Internet at http://pubs.acs.org. Received for review December 10, 2008.

November

1,

2008.

Accepted

AC802305C (62) Zhu, J.; Zhang, B.; Smith, E. N.; Drees, B.; Brem, R. B.; Kruglyak, L.; Bumgarner, R. E.; Schadt, E. E. Nat. Genet. 2008, 40, 854–861. (63) Bennett, M. R.; Pang, W. L.; Ostroff, N. A.; Baumgartner, B. L.; Nayak, S.; Tsimring, L. S.; Hasty, J. Nature 2008, 454, 1119–1122. (64) Mendes-Ferreira, A.; del Olmo, M.; Garcia-Martinez, J.; Jimenez-Marti, E.; Mendes-Faia, A.; Perez-Ortin, J. E.; Leao, C. Appl. Environ. Microbiol. 2007, 73, 3049–3060. (65) Cilindre, C.; Jegou, S.; Hovasse, A.; Schaeffer, C.; Castro, A. J.; Clement, C.; Van Dorsselaer, A.; Jeandet, P.; Marchal, R. J. Proteome Res. 2008, 7, 1199–1208.


1145

1H NMR-Based Metabolomic Approach for Understanding the

Recommend Documents