Abstract
Potassium was measured in the skins, seeds, and pulp of own-rooted Merlot grapes from the Columbia Basin, Washington. Commercial scale (7,892 kg) fermentations were carried out that varied the skin contact time and finished alcohol content. Comparison of treatments showed the control had statistically greater potassium concentration than the high ethanol and extended maceration treatments. Potassium was measured throughout the fermentations and was found to peak after crushing, diminish during the period of skin contact, and thereafter remained constant (up to 161 days). Comparison of grape and pomace potassium contents revealed that a small amount of potassium is extracted from the seed during fermentation but the opposite occurred in skins, which showed adsorption of potassium during the same period. Model wine extractions from grape skins showed that potassium extraction was increased by lowering the pH, suggesting that negatively charged polysaccharides in the grape skins may be forming ionic bonds with potassium. This theory was further reinforced by model wine extractions with pectic enzymatic treatment of grape skins soaked in potassium that showed greater percentages of skin potassium extraction than when soaked with potassium alone. Model wine extractions of grape skin showed that the combined effect of ethanol and potassium concentration both limited the extraction of skin potassium and increased potassium adsorbed to the skin than in original fruit. Thus, the model solution work both confirmed and demonstrated that the key factors in retaining and gaining potassium in skins were ethanol and potassium concentration in must/wine or model solution.
Potassium is the most abundant inorganic element found in grapes and wine (Berg et al. 1979). Throughout berry growth and ripening, potassium is accumulated in skins, seeds, and pulp (Iland and Coombe 1988, Rogiers et al. 2006). Of the berry tissues on a per unit fresh weight basis, the skin has the highest concentration of potassium (Coombe 1987) followed by the seed and then the pulp (Walker et al. 1998). The amount of potassium in the pulp is greater than the total in the skin and seed because the pulp accounts for 90% of berry mass (Possner and Kliewer 1985).
Potassium forms soluble and insoluble salts with the major and minor organic acids (primarily tartaric and malic) found in grape and plays a significant role in determining final wine pH (Boulton 1980). The solubility of potassium bitartrate salts is improved significantly by polyphenols (Balakian and Berg 1968) and declines with higher concentrations of ethanol and lower temperatures (Berg and Keefer 1958). Both alcohol and potassium will increase pH if the acid concentration is kept constant (Allen 1982, Boulton 1980). It is thought that extraction of excessive potassium from the skins causes high pH in red wines (Van Wyk 1977). Imbalance of potassium and pH has been demonstrated as a cause of stuck fermentations (Kudo et al. 1998). A 25 to 1 molar ratio of potassium to hydronium ions was identified as the minimum requirement for complete fermentation.
It is generally thought that potassium is extracted from each of the various tissues during fermentation, although there is little data to support this assumption. Small-scale experiments (49 kg) with Chardonnay showed that wines with longer contact times (up to 24 hr) finished with significantly higher concentrations of potassium (Test et al. 1986). Another study evaluated potassium concentration during small-scale (40 kg) fermentations of Shiraz, Cabernet Sauvignon, Muscat, and Riesling and found that the concentration increased during the first days of skin contact and then either declined or remained constant thereafter (Walker et al. 1998). Here, skin contact for the red varieties was limited to only three days. Two known winemaking parameters impacting final potassium are pomace contact time (Mpelasoka et al. 2003) and ethanol concentration (Berg and Keefer 1958). The purpose of this experiment was to evaluate the effects of skin contact time and alcohol concentration on the extraction of potassium during duplicated commercial scale (8.7 tons, ~7,892 kg) fermentations. Considerable effort was made in an effort to evaluate which tissue (pulp, skin, seed) contributes the most potassium during the fermentation; this localization was confirmed with model wine extractions.
Materials and Methods
Reagents.
Tartaric acid, potassium bitartrate, hydrogen peroxide, and potassium standard (1000 mg/L K+ in 3% HNO3) were from Sigma-Aldrich (St. Louis, MO).
Fruit sampling.
Six 200 berry replicates were selected at random from a 9-year-old, 8 hectare own-rooted Merlot (clone 3) vineyard located in the Columbia River Basin, Washington, on the day of harvest. Samples were put in self-sealing bags, placed on ice, and transported to the research station where they were counted and weighed. Each replicate was stored in self-sealing bag at 2°C until analysis.
Winemaking.
Merlot grapes were mechanically harvested on 12 and 13 Oct 2007 and were delivered immediately to the winery in gondolas. The grapes were received into a hopper via a Delta E50 destemmer (Vaslin Bucher, Chalonnes sur Loire, France) with the crushers disengaged and then pumped through a must line with a progressive cavity pump. The yield was approximately 0.7 L per kg (170 gallons per ton). Fermentations were carried out in duplicate. After the fruit was destemmed, 30 mg/L of sulfur dioxide was added. Diammonium phosphate (DAP) and water (control: 874 L; high ethanol treatment: 208 L; extended maceration treatment: 833 L) were added the morning after receiving, and yeast was added the evening after receiving. Yeast was added according to the manufacturer’s rehydration procedure. The DAP was added to raise the yeast assimilable nitrogen to 225 mg/L. The amount of dechlorinated water added depended on the treatment. Premier Cuvee, a Saccaromyces cerevisiae bayanus strain (Fermentis, Oskaloosa, IA) (120 mg/L), was added before mixing the tank and after the DAP and water additions were complete. The tanks were mixed at a rate of 1 min per ton of fruit. Each fermentor (18,927 L capacity) held between 7,892 to 8165 kg of must. On day one of maceration, the juice was mixed from the bottom tank valve over the top of the fermentor through an irrigator for 7 min. On day two, the tank was only mixed during the morning additions, and on day three the pump-overs began. Wine was taken from the racking valve through an air pump and pumped over the top of the fermentor through an irrigator. The tank was pumped over ~1 min per ton, three times per day until the must was drained and the pomace pressed in a Willmes press (model TP15; Willmes, Lampertheim, Germany). The press cycle lasted for a total of 46 min, beginning at 150 millibar and finishing at 350 millibar. Pressure was held for 6 min, the bladder deflated, and then rotated twice before reinflating. Each replicate was monitored for Brix and temperature once a day after the morning pump-over. A hand-held digital thermometer was used to check the temperature of the must sampled from the racking valve. The same sample was analyzed with a hand-held densitometer calibrated for Brix (DMA 35N; Anton Paar, Graz, Austria). After pressing the wines were sent to barrel where they were stored at 13°C and sampled on a weekly basis up to 161 days.
Winemaking treatments.
Initially the fruit was 28 Brix and each replicate was watered back to a different Brix target. The control was watered back to 23.8 Brix with a pomace contact time of 7 days. The high alcohol treatment was diluted to 26.8 Brix with a 7-day pomace contact time, and the extended maceration treatment was diluted to 23.8 Brix with a 20-day pomace contact time. After pressing, two 3.78-L self-sealing bags of pomace (skins and seeds) were collected from the press pan and stored at 2°C.
Fruit and pomace analysis.
Six 30-berry replicates were dissected in an effort to separate the skin, pulp, and seeds of the grape. Each replicate was placed into a large weigh boat. Berries were dissected according to a previously published method (Harbertson et al. 2002) and the juice was collected in the weigh boat, the volume of which was determined using the weight of the liquid and the density as measured with a pocket refractometer (Atago, Tokyo). The pulp solids and liquids were carefully transferred to a 50-mL tube and macerated for 30 sec with a homogenizer (Fisher Scientific, Waltham, MA) to make a fine suspension. Skins and seeds were frozen with liquid nitrogen and separately ground to a fine powder with a mortar and pestle. One gram of tissue (skin, seed, pulp) was digested in a nitric acid/hydrogen peroxide (HNO3/H2O2) solution (Jones and Case 1990). Skins from pomace were separated from the seeds and reconstructed to approximate the skins from six 30-berry replicates. The digested samples were brought to a constant volume with distilled water and analyzed for total K+ by flame emission photometry using a flame photometer (model PFP7; Jenway, Essex, UK). Samples were diluted in the range of 1- to 25-fold. A four-point standard was constructed (20 to 100 mg/L) daily for the analysis of potassium from a 1000 mg/L stock solution. The linear best-fit equation correlation coefficient varied from 0.97 to 0.99 over the analysis dates. Each day a quality-control red wine (Columbia Valley, WA) was run in triplicate to ensure consistency.
pH experiment.
Dissected berry skins were separated into 20-berry replicates. Duplicated extractions were carried out in 3 mg/mL tartaric acid solutions that were adjusted to pH 1, 2, 3, and 4 with HCl and NaOH. Solutions were added to flasks at 1 mL/berry skin for a total of 20 mL per replicate to be consistent with volume estimates of whole berries. Samples were then placed on an orbital shaking table set at 50 rpm for 24 hr. Samples were vacuum filtered using a 90 mm #1 Whatman filter paper for 3 min each. The filtrate was saved for potassium analysis and the recovered skins were ground by mortar and pestle after being frozen with liquid nitrogen. The skins and filtrate were analyzed for potassium as described earlier.
Model solution extractions.
Skins from dissected berries were separated into three 20-berry replicates and extracted with the following solutions: potassium bitartrate (4, 5, and 6 mg/mL) and potassium bitartrate (4.8 mg/mL) + pectinase (0.04 μL/mL; Adex-G, DSM, Delft, Netherlands). Each solution was adjusted to pH 3.5 with concentrated HCl. Solutions were added to flasks at 1 mL per berry skin and incubated at 30°C in a heated orbital shaking water bath at 50 rpm for 5 days. A second set of model solution extractions was performed using potassium bitartrate (4 mg/mL) but varying the concentration of ethanol from 12 to 18% v:v in steps of 2% at pH 3.5. The same extraction temperature, replication number, time, mixing, and eventual analysis of the potassium in the skin and filtrate were performed.
Wine samples were prepared by pipetting 1 mL of wine into 25 mL volumetric flasks and diluting with distilled water. The samples were then analyzed for potassium using the flame photometer as described earlier.
Experimental design.
Microsoft Excel (Redmond, WA) was used for data entry and storage. Data analysis was performed with Statistica (StatSoft, Tulsa, OK). One-factor completely randomized analysis of variance was performed on the winemaking trials and model extraction experiments. For each experiment the variance from each treatment was examined and found to be similar to the other treatments. A 5% level for rejection of the null hypothesis was used for each experiment. Fisher’s least significant difference (LSD) test was used as a post-hoc comparison of means. Similar to the ANOVA, 5% (p < 0.05) was considered a minimum for significance.
The winemaking trial was set up with a control and two treatments (high ethanol and extended maceration). Each treatment was carried out in duplicate (df = 5). The harvest fruit comparison to the pomace was carried out in quadruplicate for the skins and seeds and pomace separately (df = 15). Each of the model extraction experiments consisted of four treatments and one control. For both experiments the treatments and controls were done in triplicate (df = 11).
Results and Discussion
Results showed that titratable acidity and pH measurements of the experimental lots were similar (Table 1⇓). Evaluation of ethanol concentration of the treatments showed a significant difference (p < 0.05) (Table 1⇓). The high ethanol treatment finished with a statistically significantly greater concentration of ethanol than the extended maceration and control treatments (p < 0.05). Each treatment was monitored for potassium throughout fermentation and periodically after the wine was stored in barrels (Figure 1⇓). Potassium concentration was at a maximum shortly after crushing and slowly declined for about 7 days in the control and high alcohol treatment. The concentration continued to decline slightly during the extended maceration treatment until pressing and became constant thereafter. (Note that Figure 1⇓ axes were altered to illustrate the trend more clearly. Also note that because of the water additions during winemaking, the concentration of potassium was normalized for a better comparison.) The slow decline of potassium was somewhat inconsistent with previous work (Test et al. 1986, Walker et al. 1998), which demonstrated increased potassium with skin contact periods from 12 hr to 72 hr. In this experiment, skin contact time was 168 to 480 hr, which likely led to the significantly different results. As these grapes were mechanically harvested, the berries were likely damaged during the process, which explains why the initial increase of potassium concentration was not observed as in other research (Test et al. 1986, Walker et al. 1998). More significantly, we observed a general decline of potassium concentration during the skin contact period that was not observed by other researchers, a disparity addressed later in the discussion.
Potassium concentrations of the treatments and control were comparable to the average red wine potassium concentration as reported elsewhere (Berg et al. 1979) (1230 mg/L, n = 168) (Table 1⇑). The high ethanol and the extended maceration treatments had significantly lower concentrations of potassium than the control and were significantly different from each other (p < 0.05). The high ethanol treatment had the lowest potassium concentration of all treatments. Ethanol lowers the solubility of potassium tartrate of wine (Berg and Keefer 1958), a likely explanation for this result.
Fruit and pomace were examined in an attempt to understand which grape tissue(s) contributed the most potassium to the wine. Grapes at harvest were dissected and evaluated for their potassium concentration and compared to skins and seeds of the pomace (Table 2⇓). At harvest, the pulp accounted for ~67% of the total berry potassium, the skin accounted for ~25%, and the seed accounted for the remainder. Potassium content of the skin at harvest was significantly different than the skins recovered from the pomace (p < 0.01). The harvest skins were significantly lower in potassium concentration than the pomace skins. There were no differences in the potassium skin pomace of the different winemaking treatments. Potassium content in skins at harvest was ~20% lower than that in skins of pomace in each of the treatments and control.
Seed potassium at harvest compared with pomace seed potassium from the winemaking treatments was significantly different (p < 0.05). The control extracted 50% of the seed potassium, while both the high alcohol and extended maceration treatments extracted 50% of the seed potassium. Further comparison of seed potassium and pomace seed potassium from the winemaking treatments only showed a difference between the harvest and the winemaking treatments (p < 0.05).
The average berry weight of the grapes at harvest was ~1.2 g (data not shown). Grape mass is primarily water, so the density of water can be used to estimate 1.2 mL per berry. Using this value as a basis of estimation, potentially 2.16 g/L of potassium is available for extraction into the wine. At initial crushing the juice concentration is greater and is essentially the concentration of the pulp potassium, but the concentration declines ~15 to 25% during the skin contact period. In each case ~50% of the potential potassium is present in the wine after pressing and early stages of production.
Results showed small amounts of potassium extraction from the grape seeds with no increase in potassium concentration in the extended maceration treatment, indicating that seeds contribute very little potassium to the finished wine. Skin pomace contained greater concentrations of potassium than berry skin. The effect was consistent, with each treatment containing a~20% greater concentration than at harvest. Potassium is not easily destroyed or modified chemically, so the laws of conservation can be applied to its partitioning during fermentation. The potassium that is lost during fermentation can be accounted for by the amount present in the skin pomace (Table 3⇓). When a mass balance is applied, between 99.7 and 102.3% of the potassium can be accounted for. Given all of the error inherent in large-scale experiments and analysis, the recovery is acceptable. The freeze-dried yeast contained 50 mg potassium per gram of yeast. Thus, the yeast would contribute a negligible amount of additional potassium.
We began to investigate why there was more potassium present in the pomace skin for the treatments and control than in the fruit at harvest and why the contact time seemed somewhat connected to the decline in potassium during fermentation. Model solution experiments with grape skins from the harvest sample were used in the investigation. In each experiment, the volume of solution used was consistent with the juice yield from the same number of berries so the results could be extrapolated to winemaking. The initial investigation aimed to determine skin potassium with less harsh methodologies than those that liquify solid samples in nitric acid. Skins were extracted with tartaric acid (3 mg/mL) solutions adjusted with either concentrated HCl or 1.0 N NaOH to pH 1, 2, 3, and 4. Approximately 94% of the potassium was extracted at pH 1, and as the pH increased to 4 there was a linear decline in the percentage of potassium extracted (Figure 2⇓). The best-fit equation would predict that as the pH is lowered by 1 unit, the amount of potassium extracted would increase by ~8% until the available reserve is exhausted. The 3 mg/mL tartaric acid extractions are a less destructive technique, therefore the mechanism of extraction is not cell liquification. The pH dependence of extraction suggests that the potassium is associated with an acidic functional group, such as polysaccharides present in the cell wall. The mechanism would be an ion exchange whereby a proton is exchanged for potassium as the pH is lowered. This mechanism would require that there be a negatively charged molecule present in the cell wall of the skin. Purified grape polysaccharides (arabinogalactan, rhamnogalacturonan) from grape cell walls were found to be negatively charged from pH 2 to 5 (Vernhet et al. 1996). According to the authors, the acidic functional groups responsible for this charge are carboxylic acid groups on the uronic acid portions of polysaccharides (Vernhet et al. 1996). Results found here are consistent with these previous findings, and it is likely that a carboxylic acid present on the cell wall polysaccharides is allowing the skin to behave like an ion exchange resin. It was previously reported that the intrinsic pKa for various plant pectins was pH 2.9 (Ralet et al. 2001). In extrapolating these results to ours, we would predict that half of the acidic functional groups would be negatively charged at that pH and possible for potassium holding. In our research, at pH 3 ~75% of the potassium was extracted. One explanation for this discrepancy is that the apparent pKa is increased because of differences in the structural features of the polysaccharides present in the cell wall; a likely explanation, as we are evaluating a mixture of polysaccharides present in the grape skin.
More potassium was found on the pomace skins than on fresh skins, leading us to demonstrate the potassium-binding phenomenon in a model system and to determine if polysaccharides were possibly binding the potassium to the skins. Model solutions containing potassium bitartrate and pectinase were used to evaluate the extraction of potassium from the Merlot grape skins used previously (Table 4⇓). The experiment was designed to vary the concentration of potassium in the extraction solution. Analysis of the potassium content of the skin and the remaining skin potassium from the various extractions was significant (p < 0.001). The lowest potassium tartrate solution extracted ~53% of the potassium from the skins. However, when the potassium tartrate concentration was increased, the concentration of potassium recovered in the skins exceeded that found in the control. At 5 and 6 mg/mL potassium bitartrate, the skin potassium increased by 22 and 33%, respectively. When a small amount of pectinase enzyme was included with the potassium tartrate, the percentage of potassium extracted increased to 72%. The retentates were significantly different from the original extraction solution(s) that contained potassium (p < 0.01). The corresponding differences present in the grape skin extractions were also present in the retentates. In each case, the recovery of potassium was between 97 and 100% throughout the experiment (data not shown).
The potassium concentration range used in each of the model experiments was consistent with the range observed in pulp in the winemaking trial. A solution of potassium bitartrate containing a concentration of potassium that was lower than that observed in the pulp (1 mg/berry, ~1000 mg/L juice) extracted ~53% of the skin potassium. However, when we used extraction concentrations of potassium (1.5 and 2 mg/berry, ~1500 and 2000 mg/L juice) that were equivalent to and greater than the concentrations found in the pulp (1.5 mg/berry, ~1500 mg/L juice), significantly greater concentrations of potassium were found in the skin than in the control, consistent with results from the winemaking experiment. In these cases, the potassium from the solution had partitioned from the solution onto the skins, showing that the extraction of potassium from the skins depended on the concentration of potassium already present. When a pectinase enzyme (Adex-G) was added to an extraction containing potassium bitartrate, approximately 80% of the potassium was extracted from the skins. This result is consistent with our hypothesis that the potassium is associated with acidic functional groups on polysaccharides. Approximately 3.0 and 4.6 μM of potassium per berry were bound by the 1.5 and 2 mg/berry treatments, respectively, suggesting that the same amount of negatively charged functional groups per berry would be required of the polysaccharides to adsorb the potassium. During grape ripening, it was found that there are no overall changes in cell wall polysaccharide composition, with an estimated 2.7 mg of cell wall polysaccharides per berry (Nunan et al. 1998). Polysaccharides have multiple acidic functional groups, and it is likely they bind potassium. The enzyme-cleaved portions of the polysaccharides associated with potassium likely form pectin potassium salts in solution. Pectinases are regularly used as macerating enzymes to improve juice yield in the wine and grape juice industry (Haight and Gump 1994) and in the wine industry to increase tannin extraction (Sacchi et al. 2005). The use of pectinase enzymes in the presence of grape skins could potentially increase the potassium content of the must, and in doing so could also increase the pH. There is also the potential for pectin methyl esterase enzymes to expose more acidic functional groups to bind more potassium by demethylating methoxyl functional groups. Both pectinase and pectin methyl esterase activities are at times present in an enzymatic mixture (Andraous et al. 2004). In such cases, the potential for holding more potassium in the skins would be nullified because the pectinase enzymes would free the polysaccharides from the skins.
In order to better understand why the winemaking treatment with higher ethanol concentration had lower concentrations of potassium, we varied the ethanol concentration in the extraction solution and maintained the potassium concentration within a range that would be found in juice (Figure 3⇓). The potassium content of the grape skins and 4.8 mg/mL potassium bitartrate solution were similar to the first model solution experiment (0.55 mg/berry and 1000 mg/mL potassium). Significant differences between the control and treatments were observed (p < 0.05). With 12% ethanol, ~36% of skin potassium was extracted. When the ethanol was increased to 14 and 16%, the percentage of skin potassium decreased slightly but not enough to be statistically significant. At 18% ethanol, the extraction of potassium was nearly negligible and the same as the amount present in the control skins, but significantly different from the lower ethanol concentrations (p < 0.05). Thus, as the concentration of ethanol was raised, the potassium retained in the skin increased. At 18% ethanol, the skin potassium concentration was nearly that of the control, demonstrating that ethanol in a potassium bitartrate solution could prevent the extraction of skin potassium. It is likely that ethanol lowers the skin potassium solubility and explains its lowered extraction. We observed no crystallization of potassium tartrate salts in any of the treatments, so the concentration of the initial solution was not limited by solubility. Therefore, the combined effect of ethanol and potassium concentration explains why potassium is not extracted from the skins during the fermentations and demonstrates why greater potassium concentrations were found in pomace skin than in fruit skin at harvest.
These results suggest that skin potassium may contribute to potassium in wine if the potassium from the pulp yields 1000 mg/L or less. It is not clear how variable the amount of potassium that can be bound by a particular variety is, or if the observed concentration limits for both extraction and adsorption will be the same for all varieties. It is also not clear whether potassium strictly associates with either polysaccharide acidic functional groups or grape skins as a source of nucleation for potassium crystallization. No crystals were observed in our experiments, but it is possible that there are crystals formed that are not visible to the naked eye. Research on this topic is currently underway. Results here also indicate that the pulp potassium contributes the bulk of the potassium present in the wine and may be the best predictor for wine potassium concentration.
Conclusion
Concern about skin contact has generally been that more potassium extraction from skins would increase pH, making the wine less stable. Previous work had indicated that short skin contact times (12 to 72 hr) increase wine potassium. In our experiments, the skin contact time varied from 168 to 480 hr and results showed that potassium declines during pomace contact time and can be recovered in the skin pomace. Results suggest that skins behave like an ion-exchange resin adsorbing potassium, but releasing it in exchange for protons when the pH is lowered. Results from the model solution confirm our findings that the combination of ethanol and potassium concentration limits the extraction and enhances the retention of skin potassium. Results also suggest that juices treated with pectinase while skins are present likely increase the potassium content of the finished wine and potentially increase the pH.
Footnotes
Acknowledgments: The authors appreciate support from the Washington Wine Grape Funds. The authors also thank Chateau Ste. Michelle for their participation especially Josh Maloney for exceptional winemaking.
- Received July 2008.
- Revision received September 2008.
- Accepted September 2008.
- Published online March 2009
- Copyright © 2009 by the American Society for Enology and Viticulture