Use of Microscale Fermentations in Grape and Wine Research

Grapes.

Vitis vinifera L. cv. Pinot noir grapes were from a commercial vineyard in the Northern Willamette Valley (Willakenzie Estate, OR). Fruit from two vineyard blocks planted in 1996 with Pinot noir (clones Dijon 113 and 114) was hand-harvested and used for microscale and commercial-scale fermentations. Fruit was harvested on 25 Sept 2006 and processed the same day.

Extraction of phenolics from fruit.

Triplicate random 200-berry samples collected from harvested fruit were used for extraction and analysis of phenolics as previously described (Pastor del Rio and Kennedy 2006). Berries and extracts were kept at −80°C until extraction and analysis.

Commercial winemaking.

Fermentations took place at the Willakenzie Estate winery. Grapes were divided equally into two 4.54-ton, open-top fermentors, destemmed (Delta E₂, Vaslin Bucher, Chalonnes sur Loire, France), and blanketed with dry ice with 70 mg/kg sulfur dioxide added as potassium metabisulfite. Fruit underwent 4 days of cold prefermentation skin contact at 7.2°C prior to alcoholic fermentation. Tanks were inoculated separately with Enoferm BGY (Lallemand, Montreal, Canada) and Levuline BRG (Oenofrance, Bordeaux, France) according to manufacturer’s guidelines. The pomace was mechanically punched down once daily during prefermentation and twice daily during alcoholic fermentation, with an additional pump-over daily during alcoholic fermentation. Must and wine samples were collected every 2 days following punch down except on day 4 of cold soak and frozen until analysis. Tanks were pressed dry on day 10 of alcoholic fermentation using a bladder press (Willmes, Lampertheim, Germany) and a sample was collected.

Microscale fermentor.

The fermentor design was kept as simple as possible and consisted of a 4-L jar with a Teflon-lined cap (Olshens Bottle Supply, Portland, OR), a fermentation airlock, and a food-grade high-density polyethylene screen (Rubbermaid, Fairlawn, Ohio) (Figure 1⇓). To prevent foam overflow, fermentors were not completely filled. The cost per fermentation vessel was less than $5.00 US.

• Download figure
• Open in new tab

Figure 1

Design of the optimized microscale fermentation vessel.

Microscale winemaking.

Ten fruit clusters were randomly collected from each harvest bin used for commercial fermentations. Grapes were kept at 5°C overnight. Fruit was then destemmed by hand and divided uniformly into three fermentors. Berries (3.5 kg) were crushed using a hand-operated crusher (Mini 40 x 40, Marchisio S.P.A., Pieve di Teco, IM, Italy). The jars were blanketed with dry ice and sulfur dioxide was added at a rate of 70 mg/ kg. Musts were inoculated with Lalvin EC1118 yeast (Lallemand) after 4 days of cold prefermentation skin contact at 7.2°C, maintained with small daily additions of dry ice to minimize oxygen ingress, following the manufacturer’s guidelines. The fermentation temperature was maintained between 25 and 30°C by adjusting the room temperature. The fermentation cap was kept submerged by the plastic screen. Must density and temperature were monitored daily (DMA 35N, Anton Paar, Graz, Austria) through the airlock opening. Must and wine were sampled every 2 days from the center of the fermentation vessels through the airlock opening using a 50-mL plastic syringe fitted with a 40-cm polypropylene tube. The samples were stored at −80°C until analysis.

Wines were pressed 14 days after inoculation. The press consisted of a 6000-mL Buchner funnel fitted with a #9 stopper and a 2000-mL Erlenmeyer filtration flask (VWR International, Buffalo Grove, IL). The pomace was poured into the funnel, covered with a 0.39-mL clear, high-density, food-grade plastic sheet (Pactiv, Lake Forest, IL), and the sheet was secured with a rubber band around the Buchner funnel. Each sample was pressed three times by applying a vacuum of 1.7 bar to the Erlenmeyer flask for 2 min. The pomace was stirred between pressings. A 100-mL sample of pressed wine was collected for final wine analyses and the remainder was transferred to 2-L bottles.

Chemicals.

All solvents were HPLC-grade and purchased from various sources: phloroglucinol, methanol, ascorbic acid, and ammonium phosphate monobasic (J.T. Baker, Phillipsburg, NJ), (+)-catechin (Fluka, Buchs, Switzerland), caffeic acid (Sigma, St. Louis, MO), malvidin 3-glucoside chloride and quercetin (Extrasynthese, Lyon, France), phosphoric acid (Fisher-Scientific, Fair Lawn, NJ), hydrochloric acid, acetonitrile, and glacial acetic acid (E.M. Science, Gibbstown, NJ), and sodium acetate anhydrous (Mallinckrodt, Phillipsburg, NJ). Water for solutions was purified to HPLC-grade using a Millipore Milli-Q water system (Bedford, MA).

Instrumentation.

An Agilent model 1100 HPLC (Palo Alto, CA) consisting of an autosampler, quaternary pump, vacuum degasser, diode array detector, and column heater and coupled with Chemstation software was used for chromatographic analysis.

Chemical analysis.

Low molecular weight phenolics were analyzed by reversed-phase HPLC as previously described (Lamuela-Raventós and Waterhouse 1994). Concentrations of total and individual proanthocyanidins were determined by acid-catalyzed cleavage in the presence of phloroglucinol (phloroglucinolysis) as previously described (Kennedy and Jones 2001, Kennedy and Taylor 2003, Pastor del Rio and Kennedy 2006). In must and wine samples, a volume ranging from 24 to 6 mL (from the beginning to end of fermentation, respectively) was concentrated at 38°C to remove ethanol (centrivap concentrator, Labconco, Kansas City, MO) and then applied to a 1-g C₁₈-SPE column (Alltech, Deerfield, IL). The proportion of seed and skin proanthocyanidin extracted into wine was calculated as previously described (Peyrot des Gachons and Kennedy 2003) by comparing the wine with the pooled average of the grape skin isolates. Wine color density and hue were determined as described (Somers and Evans 1977). Volatile acid was determined by cash still distillation and expressed as g/L acetic acid.

Statistical analyses.

Statistical analyses were performed with JMP (version 6.0, SAS Institute, Cary, NC) using the standard least squares procedure for unbalanced data. Means were separated with LSD student’s t-test. The coefficient of variation (CV) was calculated as the relative percentage of the standard deviation (Excel 2003, Microsoft, Redmond, WA).

Fermentation.

Because extraction rate is temperature-dependent, attempts were made to maintain a consistent temperature across microscale and commercial fermentations. During the prefermentation cold soak, the commercial-scale fermentations were unable to reach the target temperature because of cooling inefficiency. Otherwise, fermentation temperatures of microscale and commercial fermentations were generally consistent (data not shown). The microscale fermentation temperatures were easily controlled through room temperature manipulation (±0.5°C). Although the yeast used in microscale and commercial fermentations was different because of unanticipated protocol changes, the rates of sugar depletion were very similar (Figure 2⇓). By day 6, the microscale fermentations were complete.

• Download figure
• Open in new tab

Figure 2

Fermentation progress of research and commercial ferments, assessed by specific gravity. Error bars are ± SE. Days C1 to C4 represent the cold soak period.

Phenolic extraction during fermentation.

The concentration of total flavan-3-ols (proanthocyanidin and monomers) during the course of cold soak and fermentation differed between commercial and microscale fermentations (Figure 3⇓). The first sample on day 2 of cold soak had a higher phenolic concentration than the microscale fermentors. This difference increased during fermentation. If the wines had been pressed on day 8, the phenolic concentration in the microscale fermentations would have been 50% of the commercial fermentations. To increase phenolic extraction, microscale fermentations underwent an additional four days of postfermentation maceration.

• Download figure
• Open in new tab

Figure 3

Total proanthocyanidin accumulation in research and commercial ferments, determined by phloroglucinolysis. Flavan-3-ol monomers were not subtracted from the total proanthocyanidin. Error bars are ± SE. Days C2 to C4 represent the cold soak period.

The maximum proportion of skin proanthocyanidins in microscale wines occurred six days after the commercial wines (data not shown). However, the maximum skin proanthocyanidin concentration was similar between the two winemaking methods (Figure 4A⇓). The extraction of skin proanthocyanidins over time paralleled the accumulation of total red pigments (Figure 4B⇓). Although previous research showed that tannin and red pigment extraction do not follow the same pattern (Ribéreau-Gayon 1970), the present study indicates that when skin and seed proanthocyanidin extraction are monitored separately, skin proanthocyanidin extraction does track total red pigment extraction. Interestingly, the microscale fermentations had a late “spike” in total red pigment and skin proanthocyanidin, which may have resulted from the longer maceration time enhancing grape skin tissue breakdown.

• Download figure
• Open in new tab

Figure 4

Change in (A) skin proanthocyanidin concentration, (B) total red pigments, and (C) seed proanthocyanidin concentration in microscale and commercial scale ferments. Flavan-3-ol monomers were not subtracted from the total proanthocyanidin. Error bars are ± SE. Days C2 to C4 represent the cold soak period.

The largest difference in extraction dynamics between fermentation types occurred with seed phenolic material (Figure 4C⇑). Substantial extraction of flavan-3-ols from seeds started after day 4 in commercial-scale fermentations (coinciding with completion of fermentation) but occurred four days later in microscale wines. Even with extended maceration, the microscale wines contained less seed phenolic material than commercial wines.

The rate and extent of phenolic extraction from grape berries into wine depends on the concentration of phenolics within the berry and on the maintenance of a concentration gradient. In the present study, mechanical punch-down facilitated extraction in commercial fermentations by maintaining a concentration gradient. In submerged-cap microscale fermentations, the concentration gradient is reduced because the cap is not agitated. Punch-downs also contribute to physical breakage, releasing seeds from inside the berry and increasing their concentration gradient. Very few seeds were seen in the microscale fermentors, indicating that there was less physical berry disruption than in commercial fermentations. After day 8 of fermentation, berry degradation occurred in the microscale fermentation and the seed proanthocyanidin concentration increased.

Phenolic profile of finished wines.

The flavonol concentration of commercial-scale wines was slightly lower, but no differences in concentrations of hydroxycinnamic acids, gallic acid, or stilbenes were observed (data not shown). The concentration of flavan-3-ol monomers was higher in commercial wines, consistent with higher seed phenolic extraction as indicated by phloroglucinolysis (Table 1⇓). Commercial wines also had less total anthocyanins and slightly less bisulfite-resistant pigment (data not shown). It is possible that increased incorporation of anthocyanins into polymeric species occurred in commercial wines, possibly because of increased oxygen availability. Oxygen and oxidation products participate in anthocyanin and other polyphenol polymerization reactions (Mirabel et al. 1999, Dallas et al. 2003). Also, pigmented polymers can undergo bisulfite bleaching and anthocyanin-containing reaction products can be colorless (Cheynier et al. 2006).

As determined by phloroglucinolysis, the final total proanthocyanidin concentration was ∼19% lower in microscale wines, with no observed difference in mean degree of polymerization (Table 2⇓). The difference between proanthocyanidin extension and terminal subunit molar proportions again reinforces the observation that more seed extraction occurred in commercial wines. Skin proanthocyanidin proportion was ∼10% higher in microscale wines, with no difference in total concentration (Table 3⇓).

Compared with commercial fermentations, microscale wines had increased color intensity (520 + 420 nm) and no difference in hue (420/520 nm) (Table 4⇓). Others have observed that phenolic oxidation increases hue (Singleton 1987, Cheynier et al. 1990, Castellari et al. 1998). If the fermentations were not protected against oxygen, higher hue values would be expected in microscale wines. The hue values reported in this study are lower than those found in other studies on Pinot noir wines (Mazza et al. 1999, Reynolds et al. 2005), suggesting that our method effectively minimized oxidation. Protection against oxidative spoilage was also confirmed by the lower volatile acidity of microscale wines.

Method potential.

Large-volume fermentations are usually not feasible in research because of limitations in fruit production, facilities, equipment, and cost. Large-scale fermentations are also more difficult to manage and control, as seen in this study with temperature. This can result in confounding errors and inconsistent research findings.

Differential extraction of skin and seed phenolic components presents an impediment to predicting wine composition based on fruit analysis. Several authors have attempted to develop methods of berry extraction that mimic a wine fermentation (Singleton and Draper 1964, González-Manzano et al. 2004, Canals et al. 2005, Sun et al. 2006). These methods usually involve a model wine system varying in alcohol concentration and pH. Extraction into a model system presents many drawbacks; principal among these is that the extraction system is not maceration under fermentation conditions.

The method presented here integrates the scale of a model extraction with the dynamics of fermentation and is potentially a better option for evaluating the impact of viticultural practices on wine composition. While ultimately the final extraction and composition of wine is under winemaker control, microscale wine production provides the option of producing wines in a highly reproducible way and can be adapted to reproduce individual winemaking conditions.

Variability across microscale fermentation replicates was very low. The coefficient of variation (CV) was always less than 0.5% for must density and temperature monitoring during fermentation, while low molecular weight phenolic compounds had an average CV of 2% (data not shown). Proanthocyanidin analysis and spectral evaluation of the final wines revealed a variability of 1.5–2.5% of the mean among replicates, much lower than the relatively low 5–9% variability among commercial tanks due to the homogeneity of vineyard blocks.

Although this technique did not achieve the same extraction profile during the course of fermentation, particularly for seed-derived phenolics, modifications of the microscale method could be made to more closely resemble commercial wines. For example, having a higher fermentation temperature would likely accelerate extraction of seed material. Also, since seed extraction becomes more important when ethanol is present, a longer maceration would probably increase total tannin and the proportion of proanthocyanidins derived from the seed (Peyrot des Gachons and Kennedy 2003, Canals et al. 2005). This microscale method could be modified to accommodate different cap management (punch-down) while still maintaining an anaerobic environment. In this study, however, it was thought that minimizing wine production variables by having a submerged cap would result in a more reproducible fermentation. The microscale fermentation method has been used successfully to produce wines from a large rootstock trial where only small amounts of fruit from each rootstock were available, yielding enough wine for sensory testing (Sampaio et al. 2006).