Abstract
Microbial analyses of older and recently bottled wines demonstrated the ability of certain wine microbial species to survive and grow in the bottle. Among them, the spoilage yeast Brettanomyces bruxellensis was predominant, necessitating effective methods for removing these microorganisms before bottling. This work reports on various filtering assays. Each technique was evaluated over several years after bottling by microbial analysis and by volatile phenols measurements. The smaller the pore size, the more microbes were eliminated. Elimination of bacteria required a 0.3-μm filter, but a 1.0-μm filter was efficient for yeast elimination. In more tightly filtered wines, volatile phenol concentrations were lower than in less tightly filtered wines and in nonfiltered wines.
Microorganisms influence sensory properties of wine by producing many secondary products such as glycerol, esters, and alcohols (van Vuuren and Dicks 1993, Romano et al. 2003). Following fermentation, sulfur dioxide is added to reduce the microbial population and to prevent growth of spoilage microorganisms. Some metabolites can dramatically change a wine’s sensory composition, most notably volatile phenols produced by Brettanomyces bruxellensis (Chatonnet et al. 1995). Moreover, some bacterial species can produce biogenic amines (Coton et al. 1998) or ethyl carbamate (Uthurry et al. 2005), which are undesirable to consumers. During barrel aging, wine cannot be totally sterilized, since each enological operation is a possible source of contamination. At the end of aging, the wine microbial community is stabilized but the total microbial population in the wine before bottling is often 103 to 104 viable and culturable cells per milliliter. Residual micro-organisms are mainly Acetobacter aceti and the residual fermentative species Saccharomyces cerevisiae and Oenococcus oeni. Some spoilage species such as Pediococcus parvulus and B. bruxellensis also remain.
Microbial ecology studies have focused on primarily microbial changes made during winemaking and aging (Renouf et al. 2006a,b). However, for the majority of red wines from Bordeaux, the duration of fermentation and of aging is much shorter than storage time in the bottle. Previous microbiological studies have been made at each step of bottling (Donnelly 1977, Neradt 1982). Similar studies have been made on bottles, cork, and screwcaps (Costa and Pereira 2005). Recently, new bottling equipment, sanitation procedures, and the use of hazard analysis and critical control point (HACCP) systems have contributed to the reduction of microbial populations and contamination at the bottling stage. Nevertheless, some indigenous microorganisms may remain in a wine after bottling.
The goals of this study were to determine the populations of residual microflora at the end of aging and their preservation during storage in the bottle, to examine the most common prebottling filtration methods as means of eliminating bacteria and yeasts, and to inventory the yeast and bacterial species detected in older wines bottled without prebottling filtration.
Materials and Methods
Wine samples.
Wines were collected from various wineries in different areas of the Bordeaux appellation: Saint-Emilion, Pessac-Léognan, Margaux, and Pauillac. For all wines, three bottles were analyzed and averages and standard deviation were calculated. All samples studied were conserved under similar conditions, particularly temperature and humidity, and were neither diluted nor concentrated before microbial and chemical analyses.
Recent vintages.
Three wines of the 2003 vintage from Saint-Emilion, Pessac-Léognan, and Pauillac were monitored during their development until bottling, which did not include filtration. Chemical and microbiological data at bottling are reported in Table 1⇓. These wines were analyzed after 1, 3, 6, and 10 months in bottle. Three bottles were analyzed for each group.
Filtration.
Filtration experiments were conducted in two cellars and repeated over three successive vintages. Experiments on the 1996, 1997, and 1998 vintages were performed in a Saint-Emilion cellar, using K700 (7.0 μm) and K300 (4.0 μm) (ScottLab, Petaluma, CA) filters. Experiments on the 1993, 1994, and 1995 vintages were performed in a Pauillac cellar, using K900 (10.0 μm), K300 (4.0 μm), K100 (1.0 μm), and EK (0.3 μm) filters. Bottles of each wine were analyzed in triplicate in 2006. Data from filtered wines were compared to nonfiltered wines.
Older vintages.
Ten older vintages of Bordeaux wines (from 1909 to 1981) were studied in triplicate (Table 2⇓). None had been filtered before bottling. These wines were of marketable quality and were not spoiled by apparent taints or cloudiness. When they were opened in 2006, their corks (retained since bottling) were in good condition. All bottles were sealed with wax.
Microbial analysis.
Wine samples were analyzed by direct epifluorescence to determine the viable microflora. Driven equilibrium Fourier transform (DEFT) was carried out with a Chemunex system (Ivrysur-Seine, France) using 10 mL of wine sample (Millet and Lonvaud-Funel 2000).
Wines were subjected to microbial analysis to determine populations of total yeast (TY), non-Saccharomyces yeast (NS), lactic acid bacteria (LAB), and acetic acid bacteria (AAB). Yeast growth medium contained 10 g/L yeast extract, 10 g/L bactotryptone, 20 g/L glucose, and 20 g/L agar. The pH was adjusted to 5.0 with orthophosphoric acid. The medium for cultivation of total yeast was supplemented with biphenyl (0.015% w/v) (Fluka, Steinheim, Switzerland) and chloramphenicol (0.01% w/v) (Sigma Aldrich, Saint-Quantin, France) to inhibit mold and bacterial growth. The addition of 0.025% (w/v in acetone) cycloheximide (Sigma-Aldrich) prevented Saccharomyces spp. growth and allowed enumeration of the non-Saccharomyces yeast population. Incubation at 25°C lasted 5 days for total yeast and 10 days for non-Saccharomyces yeast. For bacteria, the medium consisted of 250 mL/L commercial red grape juice, 5 g/L yeast extract, 1 mL/L Tween 80, and 20 g/L agar. The pH was adjusted to 5.0 with 10 N KOH. Yeast and mold development was inhibited by adding 50 mg/L pimaricine (Delvocid, DSM Food Specialities, Delft, Netherlands). Enumeration of LAB population was accomplished via anaerobic incubation using an anaerobic system envelope with palladium catalyst for 7 days at 25°C. For AAB population enumeration, the same medium was supplemented with 30 mg/L penicillin (Sigma Aldrich) to prevent Gram-positive bacteria growth. Incubation was 3 days at 25°C. Assays were performed in triplicate and results were expressed as colony forming units (cfu/mL).
Three bottles were analyzed for each assay, one from the beginning of the bottling line, the second from the middle, and the third from the end. Each bottle was then analyzed in triplicate. For each assay, the microbial populations varied from one bottle to another, generally by 10% but sometimes by more. The wide variation observed resulted from this analyses repetition and use of bottles throughout the bottling line as replicates. The repetitions of the samples and of their treatment are essential to be able to consolidate the differences observed and to treat them statistically.
Microorganism identification.
Yeast identification was made through use of molecular methods. DNA was extracted from 100 mL of wine using a previous DNA extraction protocol (Renouf et al. 2005a). Identification was based on sequence analysis of D1/D2 domains of the rRNA 26S gene by polymerase chain reaction-denaturing gradient gel electrophoresis (PCR-DGGE) (Cocolin et al. 2000). Each suspect band of the acrylamide gel was excised for further sequencing. Sequences were identified by alignment and phylogenetic comparison (Saitou and Nei 1987) with authentic sequences from GenBank. For yeast species identification, we also used restriction fragment length polymorphism (RFLP) analysis of the 5.8S rRNA gene and the two ribosomal internal transcribed spacers (ITS1 and ITS2) (Esteve-Zarzoso et al. 1999) on isolated colonies from TY plates to provide the percentage of each culturable yeast species (Table 3⇓).
Brettanomyces bruxellensis was identified based on colonies isolated from NS plates and by the species-specific nested-PCR method (Ibeas et al. 1996) with the DB1/DB2 primer set. The PCRs were performed directly on colonies. An initial stage of cell lysis at 95°C during 5 min preceded the PCR amplification program.
For bacteria, analyses were made on the whole biomass collected after centrifugation (20 min, 10,000 x g, 4°C) of 100 mL wine. Then DNA was extracted and analyzed by PCR-DGGE targeting the rpoB gene (Renouf et al. 2006b). Oenococcus oeni colonies isolated on LAB plates were tested by a species-specific PCR method for O. oeni (Divol et al. 2003) performed directly on the colonies, which gave their percentage in the LAB population.
Chemical analysis.
Conventional analyses of pH, total acidity, volatile acidity, alcohol content, free and total SO2, and total polyphenol index were carried out by the official methods or the usual methods recommended by the International Organization of the Vine and Wine. Malic acid, glucose, and fructose concentrations were measured by the enzymatic method (Biopharm, Darmstadt, Germany). Volatile phenols were extracted with dichloromethane from a 50-mL sample and quantified by gas chromatography (Chatonnet and Boidron 1988).
Statistical analysis.
The effect of different factors on microbial populations was analyzed using SigmaStat (two-way ANOVA test) (Systat Software, San Jose, CA). A probability (p) of less than 0.05 indicated that the variable under consideration had a significant effect on the population level.
Results
Microbial changes in newly bottled wine.
DGGE NL1/LS2 profiles were obtained for wine P during the first 10 months spent in bottle (Figure 1⇓). At bottling, three species were detected: S. cerevisiae, H. uvarum, and B. bruxellensis. After bottling, H. uvarum was slightly detectable after 1 month and S. cerevisiae was only occasionally detected. After 6 months, B. bruxellensis was the only species detected. There was no difference between enumerations on plates and epifluorescence analysis. In all wines, the AAB population decreased quickly in bottle. LAB populations increased in Pauillac (3.3 × 105 cfu/mL) and Saint-Emilion (1.1 × 104 cfu/mL) wines and decreased in Pessac-Léognan wine (<1.0 cfu/10 mL). In Pessac-Léognan, the TY population reached 2.5 × 104 cfu/mL and the volatile phenol concentrations rapidly increased (Figure 2⇓), exceeding the critical olfactory threshold of 400 to 600 μg/L (Suarez et al. 2007). After 10 months storage in bottles, volatile phenol concentrations increased significantly in all wines (Table 4⇓), with Pessac-Léognan having the highest and Pauillac wine the lowest. In Pauillac, the yeast population remained stable (1.6 × 103 cfu/mL) unlike the LAB population, which was predominated by P. parvulus species (O. oeni represented only 10% of colonies from LAB plates).
Filtered and nonfiltered wines.
Microbial and volatile phenol concentrations of wines analyzed for the filtration assays are reported in Tables 5⇓ and 6⇓. No AAB population was found. Most nonfiltered wines contained high TY and LAB populations. There was no significant difference between the plate counts and the viable populations counted by epifluorescence. B. bruxellensis was the main yeast detected in these wines. Residual S. cerevisiae and P. anomala species were detected, but in very low proportion (<5% of colonies from TY plates by PCR-RFLP analysis). O. oeni was the only bacteria species detected in Saint-Emilion wines. In Pauillac wines, P. parvulus was detected in addition to O. oeni (Figure 3⇓).
Statistical analysis was performed to compare the effects of the different filtrations. For wines made in 1996, 1997, and 1998, filtration method had no statistically significant effect on LAB population. Filtration impact on yeast population varied according to vintage. In 1996 and 1998 wines, the K300 filter significantly reduced TY populations (p = 0.003, p = 0.004). It also statistically reduced NS populations (p = 0.002, p = 0.002) and volatile phenol concentrations (p = 0.001, p = 0.001). These effects were not statistically significant for the 1997 vintage (p = 0.423 for TY population), but this vintage had the lowest yeast population and the highest ethanol concentration (% vol).
The second series of assays on Pauillac wine showed the impact of K100 and EK filtration (Table 6⇑), which eliminated total yeast for each vintage and prevented the increase of volatile phenols. In 1994 wines, K300 filtration was sufficient to eliminate the yeast population but there was only partial reduction in 1995. LAB populations were never detected in wines filtered by EK sheets. The K100 filtration was sufficient to eliminate LAB population only in 1994. For this vintage, O. oeni was the only species found. For other vintages, P. parvulus was also detected. These species were detected in nonfiltered and K100-filtered wines.
Microbial community in older wines.
The great majority (90%) of older bottled wines contained high yeast populations (Table 7⇓). Wine B bottled in 1909 contained 2 × 103 cfu/mL viable and cultivable non-Saccharomyces yeasts in 2006. Only 40% of the older bottled wines contained LAB, but in those that did, the population was often high. For example, wine F bottled in 1949 contained a LAB population of 4.1 × 106 cfu/mL. AAB was not found in any of the older wines. These microorganisms counted could be considered viable and cultivable since no significant difference was observed between plate counting and epifluorescence.
Five yeast and four bacteria species were respectively identified by PCR-DGGE after comparison and phylogenetic analysis. Wines F and J had the highest LAB population, as shown in DGGE gels (Figure 4⇓). The most frequently detected species was O. oeni, which was present in wines D, G, and J, and it represented more than 90% of the colonies isolated on LAB plates. The other frequently detected species detected was P. parvulus. In wine F, which had the highest LAB, O. oeni was not detected but P. parvulus and P. damnosus were detected.
The main yeast species was B. bruxellensis, and it was present in all wines in which cultivable yeasts were counted. For half the wines (A, B, D, H, and J), B. bruxellensis was the only yeast species detected. All these wines contained high concentrations of volatile phenols.
Discussion
In the bottle, the absence of oxygen and low concentrations of potential substrates limited growth for most species, and the species diversity was obviously less than on grape (Renouf et al. 2005a), during fermentation (Renouf et al. 2005b), or during aging (Renouf et al. 2006a), where environmental conditions were more favorable. The main wine yeast, S. cerevisiae, was detected in older wines but in very low proportion. AAB, for which the presence of oxygen was essential, was never detected in the bottled wines. Therefore, non-Saccharomyces yeast and lactic acid bacteria were the main species surviving in bottled wines. Survival of LAB populations could be explained by the fact that malolactic fermentation was not carried out in old Bordeaux vintages. Among the LAB, O. oeni and P. parvulus were most often detected.
Of the non-Saccharomyces yeast, Pichia anomala has been detected on grape surfaces and is active at the onset of alcoholic fermentation (Cocolin et al. 2001, Renouf et al. 2006c). Its presence in some bottled wine could be explained by its low survival requirements at homeostasis (Fredlund et al. 2002). Another yeast species occasionally detected in bottle is Zygosaccharomyces bailii. This species could be more problematic because of its ability to referment sweet bottled wines (Loureiro 2000, Divol and Lonvaud-Funel 2004). The main species in bottled wine was B. bruxellensis. Its predominance in older wines could be attributed to its exceptional survival characteristics under minimal nutrient conditions (Uscanga et al. 2000) and to its ability to use alcohol as a carbon source.
The presence of active microorganisms in bottled wine was equated with loss of quality, emphasizing the importance of removing them at bottling to ensure wine stability during storage. Filtration could eliminate solid particles, which include microbial cells. Microbial retention depended on the size of cell and the grade of the filter sheet. Clearly, the smaller the filtering pores, the more effectively the microbes were removed. However, very fine filters could also reduce wine aromatics and color characteristics (Flores et al. 1991). Thus, it is crucial to find the best possible compromise between microbiological stabilization and sensory impact. The ideal pore size should result in minimum populations of microbes while not affecting sensory quality.
Filtration assays described here demonstrated that EK filtration was required to remove all bacteria but K100 filtration was sufficient to remove all yeast. Consequently, the choice should be made after a prebottling microbial analysis combining enumeration on plates for culturable populations and epifluorescence for estimation of nonculturable cells. When the bacteria population was lower than the yeast population, K100 filtration was sufficient to stabilize the wine. When the reverse is the case, and notably if Pedioccocus spp. are detected, then EK filtration is recommended.
Prebottling filtration should be considered an ultimate step when traditional practices such as sulfiting, racking (Renouf and Lonvaud-Funel 2004), and fining (Murat and Dumeau 2003) are unsatisfactory. Other prebottling operations (thermal treatment and use of antimicrobial agents) should also be considered for further study.
Conclusion
Certain non-Saccharomyces yeast and lactic acid bacteria species survived in the bottle for very long time. Some species disappeared during aging and reappeared during bottle storage, demonstrating their ability to acquire progressively strong resistance, which could allow passage of a viable but nonculturable, very low population at the end of the winemaking process, below the detection threshold. Analyses of bottled wines showed no significant difference between plate count values and the viable cells observed by epifluorescence. Among bacteria, O. oeni was found in young wines, sometimes with Pediococcus spp. Among the yeasts, B. bruxellensis was predominant and could grow during bottle storage over a long period, which could produce 4-ethylphenol and 4-ethylguaiacol concentrations exceeding critical olfactory thresholds during the first months of storage.
Filtration, regarded by some winemakers as responsible for stripping wine fruit character and reducing body and viscosity, appeared to be the best solution to prevent microbial contaminations. A 1.0-μm grade filter sheet was sufficient to eliminate all yeasts, and it significantly prevented increased volatile phenols concentrations over several years after bottling. For bacteria, a 0.4-μm grade filter sheet was necessary. The decision between filtration and/or other stabilizing practices should be determined by microbial analysis before bottling, combining use of plates for enumeration of culturable populations, epifluorescence for estimation of viable but nonculturable cells, and molecular tools for species identification.
Footnotes
Acknowledgments: The authors thank C. Congé, J.P. Masclef, V. Millet, J.L Thunevin, and A. Vauthier for supplying grape and wine samples from their chateau. We also thank Joana Coulon and John Davis for the redaction assistance.
- Received January 2007.
- Revision received March 2007.
- Revision received April 2007.
- Copyright © 2007 by the American Society for Enology and Viticulture