Properties of Acetaldehyde Condensation Reactions with Glycerol, (+)-Catechin, and Glutathione in Model Wine

pubs.acs.org/jafc 1 H NMR: A Novel Approach To Determining the Thermodynamic Properties of Acetaldehyde Condensation Reactions with Glycerol, (+)-Catechin, and Glutathione in Model Wine Ana L. Peterson and Andrew L. Waterhouse* Department of Viticulture and Enology, University of California, One Shields Avenue, Davis, California 95616, United States *S Supporting Information ABSTRACT: As wine oxidizes, ethanol is converted to acetaldehyde, but its accumulation is not predictable, due to poorly characterized reactions with alcohols, SO 2, thiols, flavanols, and others. Measurement of these components has been thwarted by equilibria into the other forms during sample preparation. NMR spectra can be taken on intact samples and is thus ideal for this situation. Equilibria of acetaldehyde with glycerol, (+)-catechin, and glutathione were studied separately in model wine solutions at ph 3 4 by 1 H NMR and 2D ( 1 H 1 H) COSY spectra. Glycerol acetals had equilibrium constants between 1.14 ± 0.056 and 2.53 ± 0.043 M 1, whereas ethylidene-bridged (+)-catechin dimers and glutathione thiohemiacetals had more favorable equilibria: from (3.92 ± 0.13) 10 3 to (6.13 ± 0.32) 10 3 M 2 and from 10.18 ± 0.22 to 11.17 ± 0.47 M 1, respectively. These data can be used to create accurate measures of acetaldehyde in its various forms and, consequently, offer insight into wine oxidation. KEYWORDS: acetaldehyde, (+)-catechin, glutathione, glycerol, equilibrium constants INTRODUCTION Aging is crucial to the development of a red wine and occurs partly through its exposure to oxygen as first demonstrated by Pasteur via comparing wine sealed in an ampule with oxygen versus no headspace. 1 Wine oxidation is dependent on the method of aging, wine composition, and other factors. The exposure of wine to oxygen induces a cascade of radical reactions and subsequent generation of new compounds, many of which can alter the aroma, color, and even mouthfeel of wine. The interaction between oxygen and various wine components has been studied in numerous model wine experiments to determine the mechanisms by which ethanol and other substrates are oxidized. 2,3 The aroma complexity and pigmentation of a red wine can increase with modest oxygen exposure. 4 Micro-oxygenation accelerates the aging process through controlled additions of oxygen; however, metrics to monitor the process are lacking, but necessary to gain better control over the changes and thus wine quality. Currently, aged and micro-oxygenated wines are monitored in production by sensory analysis or, in some research situations, by measuring acetaldehyde. 5 7 The electrophilicity of acetaldehyde leads to its participation in reversible reactions with various nucleophiles in wine, including alcohols, thiols, sulfur dioxide SO 2, and flavonoids. Consequently, its measurement is complicated by the formation of these adducts, and measurement of free acetaldehyde does not reflect the total amount present or the amount of oxidation that has occurred (Figure 1). In an aqueous solution, acetaldehyde rapidly (acid is a catalyst) 8 equilibrates with water, forming the hydrate (4), 9 and a parallel reaction with ethanol yields the hemiacetal (3a) and acetal (1,1-diethoxyethane, 3b). Diethyl acetal has been reported to have a fruity character in commercial sherry wines undergoing biological aging. 10 Four heterocyclic acetal isomers are produced from glycerol, cis- and trans-dioxolane (1a,b) and cis- and trans-dioxane (2a,b), and have been identified as oxidation markers in fortified wines, the amount correlating with barrel age in Madeira and Port wines in which slow oxidation is ongoing. 6,11 Acetaldehyde reacts with flavan-3-ols and anthocyanins, altering the overall color and mouthfeel (astringency) 12 with condensation and polymerization reactions. 13 Bridging reactions between flavanoids (i.e., 5) have been studied in real and model wine solutions. 14 16 These reactions are reversible, and one method has been described to recover the acetaldehyde from these products. 7 Acetaldehyde can also act as a nucleophile and attack the electrophilic portion of the flavylium cation (anthocyanin), forming a pyran ring and yielding vitisin B, (6). 17 Addition of SO 2 to wine for the prevention of oxidative and microbial spoilage results in an equilibrium between various S(IV) species, bisulfite (HSO 3 ) being the predominant form at wine ph. Bisulfite can also undergo nucleophilic addition with carbonyl compounds in wine, yielding hydroxyalkylsulfonic acids (7). The dissociation constant for bisulfite-bound acetaldehyde under wine-relevant conditions is well documented and indicates that the sulfonate product is highly favored. 18 The sensitivity of wine consumers to SO 2 has led to investigations into the use of glutathione as an alternative antioxidant, specifically its ability to scavenge quinones under wine-like conditions. 19 The addition of this thiol nucleophile to Received: May 6, 2016 Revised: August 9, 2016 Accepted: August 13, 2016 Published: August 31, 2016 2016 American Chemical Society 6869

Figure 1. Molecular depiction of some of the equilibria formed between acetaldehyde and its various derivatives: trans- and cis-dioxolane (1a and 1b), trans- and cis-dioxane (2a and 2b), ethyl hemiacetal (3a) and diethoxyethane (3b), acetaldehyde hydrate (4), ethylidene-bridged catechin dimer (5), pyranoanthocyanins (vitisin B, 6), hydroxyethylsulfonic acid (7), and the glutathione adduct (8). acetaldehyde yields a thiohemiacetal (8). 20 Thus, the glutathione thiohemiacetal is another possible pathway for acetaldehyde consumption. Gas chromatographic methods for acetaldehyde involve solid-phase microextraction, 21 followed by derivatization with PFBHA. 22 Liquid chromatography methods use DNPH 23 or cysteamine. 24 The addition of heat and/or changed ph for derivatization methods will catalyze the hydrolysis of acetals 25 and cleavage of ethylidene bridges 7 as well as hydrolyze hydroxysulfonates, 23 altering free acetaldehyde. These procedures will therefore create interferences for measurements of free acetaldehyde. Although many addition products of acetaldehyde have been studied in wine, the concentration of free acetaldehyde is not usually reported. 26,27 To determine the equilibrium point of these reactions, the concentration of all components must be known. Because the available methods likely measured some of the addition products, but at best were tested against only one of the addition reactions, it was necessary to find a method for the analysis of reaction solutions that did not disrupt or shift the equilibria and could measure all components. Although less sensitive than chromatographic methods (5.1 μg/l 28 ), NMR spectroscopy (1.1 mg/l 29 ) is capable of the simultaneous detection of many components and has minimal sample preparation and, thus, would not affect reaction equilibria. NMR has advanced its use for quantitation of components and interactions in real and model wine solutions. 30,31 The development of new pulse sequences for the suppression of water and ethanol 32 and the optimization of 2D experiments such as diffusion-ordered NMR spectroscopy (DOSY) to investigate changes in wine through aging 33 have improved the selectivity, sensitivity, and resolution of wine analysis by NMR. NMR studies of acetaldehyde in aqueous solutions have revealed the equilibrium that exists between acetaldehyde and its hydrated form. 34,35 NMR has also been used to measure the acetaldehyde equilibria with hemiacetals (acetal intermediates) 25 and, more recently, the formation of acetaldehyde hydrate as well as poly(oxymethylmethylene) glycols (hydrate polymerization products). 36 Nikolantonaki et al. successfully applied NMR spectroscopy to the analysis of acetaldehyde and other carbonyls in the free and SO 2 -bound forms in wine to overcome measurement difficulties caused by their instability and chemical reactivity. 29 This prompted our investigation and optimization of 1 H NMR for the measurement of acetaldehyde interactions with other nucleophiles in model wine solutions. Under wine-relevant conditions, 1 H NMR was used to investigate the equilibria between acetaldehyde and its derivatives in interactions with sulfites, flavonols, and thiols. MATERIALS AND METHODS Chemicals and Reagents. Acetaldehyde (99.5%) and glutathione (98%, reduced) were purchased from Acros Organics (Morris Plains, NJ, USA). Acetaldehyde diethyl acetal (99%), 3-trimethylsilylpropionic-2,2,3,3-d 4 acid sodium salt (98 atom % D), deuterium oxide (99.9 atom % D), glycerol (99.5%, ACS reagent), (+)-catechin hydrate (98%), and (+)-tartaric acid ( 99.5%) were all obtained from Sigma- Aldrich (St. Louis, MO, USA). Sodium hydroxide solution (10.05 9.95 N certified) was purchased from Fisher Scientific (Fair Lawn, NJ, USA). Deuterium-labeled ethanol (ethanol-d 6, 99% D, <6% D 2 O) was obtained from Cambridge Isotope Laboratories, Inc. (Tewksbury, MA, USA). All aqueous solutions were prepared with Milli-Q water from Millipore (Bedford, MA, USA). Reaction Solutions. All reaction solutions were made up in a tartrate buffer containing 6 g/l tartaric acid and brought to exact ph levels of 3.0, 3.5, or 4.0 with 10 N sodium hydroxide solution. Acetaldehyde was added to the buffer using a Drummond positive displacement microdispenser to give a final concentration of approximately 10 mm. From this solution 500 μl aliquots were removed and placed in 2 ml vials to which 72 μl of deuterated ethanol (d 6 -EtOD) and 50 μl of deuterium oxide (D 2 O) containing 2.05 mm deuterated trimethylsilylpropanoic acid (d 4 -TMSP) were added. These small volume solutions were made to simulate a model wine, giving approximately 11.6% ethanol in the deuterated form. Vials were vortexed for a few seconds to ensure complete homogeneity. Solutions were then transferred to Bruker 5 mm NMR tubes. Using liquid nitrogen, solutions within the tubes were freeze thaw degassed and then flame-sealed under vacuum. All heated reaction solutions 6870

were kept in water baths at the temperatures specified in data tables. For slower reactions (with glycerol and (+)-catechin), solutions were prepared in triplicate after which 1D spectra were taken of the reaction mixture every 24 48 h to monitor the progress and eventual equilibration of the reaction. For all reactions, 2D spectra were collected at equilibrium for structural confirmation. Glycerol Acetals. Glycerol was added to tartrate buffers (ph 3.5) to give concentrations of approximately 200 and 100 mm glycerol and ratios of glycerol to acetaldehyde concentrations of 20:1 and 10:1, respectively. At ph 3.0 and 4.0, only the 10:1 glycerol to acetaldehyde concentration ratio was investigated. The acetal reaction was also investigated at 25 and 50 C with the 10:1 glycerol to acetaldehyde concentration ratio and ph 3.5. Ethylidene-Bridged Catechin. Solid (+)-catechin hydrate was dissolved in the tartrate buffer (ph 3.5) to give an approximate concentration of 10 mm and ratios of (+)-catechin to acetaldehyde concentrations of 1:1 (equimolar) and 2:1 (for which the acetaldehyde concentration was decreased to approximately half the catechin concentration). Equimolar amounts of the two reactants were used for the investigation of the equilibrium at ph 3.0 and 4.0. Glutathione Adduct. Solid glutathione was dissolved in tartrate buffer to give approximate concentrations of 10 and 100 mm and ratios of glutathione to acetaldehyde concentrations of 1:1 and 10:1, respectively. Equimolar amounts of the two reactants were used for the investigation of the equilibrium at ph 3.0 and 4.0. NMR Spectroscopy. All experiments were performed on a 600 MHz Bruker Avance spectrometer, equipped with a 5 mm TCI cryoprobe. The 1D 1 H NMR spectra (used for quantitation of reaction components) were acquired with the standard pulse sequence, zgpr, which suppresses the water signal at 4.69 ppm by presaturation. The power for presaturation (p19) was set at 40 db, and the spectral width was 7183.9 Hz. Data were collected into 65,536 data points over 32 scans and 4 dummy scans. The acquisition time and relaxation delay were 3.0 and 5.0 s, respectively. The 1 H chemical shift homonuclear couplings for each reactant studied were determined by 1D 1 H NMR and homonuclear 1 H 1 H correlation spectroscopy (COSY) analysis of model wine solutions (prepared as indicated above) containing reactant standards. Reaction solutions were also analyzed by 2D-COSY to confirm the presence of products and elucidate possible side products and intermediates. Spectra were acquired with a COSY pulse sequence with presaturation of the water signal (cosygpprqf) at 4.69 ppm. The time domain data matrix was 4096 (F2) by 256 (F1) with 16 or 32 transients obtained per increment, depending on the sensitivity necessary. The spectral width was 7215 Hz in both dimensions. The relaxation delay was set to 2.0 s. All spectra were calibrated according to the d 4 -TMSP signal at 0.0 ppm. 37 Spectra were processed by zero filling and Fourier transform to 8K by 2K data points after multiplication with sine filter. Quantitation of Reaction Components. 1 H NMR was used to quantitate acetaldehyde, glycerol, and the resulting glycerol acetal products every 10 40 h, depending on the conditions and thus rate of reaction. Quantitative 1 H NMR acquisition parameters were as reported in the section above. TMSP served as a chemical shift standard as well as a calibration standard. Representative signals for each compound of interest were chosen on the basis of resolution. The integrations of these signals (peak areas) were converted to concentrations by comparison with integration of the calibration standard signal, for which the concentration was known. The molar ratio of the analyte to internal standard (d 4 -TMSP) is given by the expression M M analyte TMSP Aanalyte N = A N TMSP TMSP analyte where M, A, and N are molarity, peak area, and number of nuclei, respectively. 38 From this expression, the molarity of the analyte present can be determined through integration of the analyte and TMSP signals. Peak areas for quantitation of compounds were acquired by peak picking in combination with line-fitting deconvolution (using Global Spectral Deconvolution) on the Mestrenova 9.0 software. Calculation of Equilibrium Constants. Apparent equilibrium constants were calculated as the division of the equilibrium concentrations of the product(s) by the product of the equilibrium concentrations of the reactants. [acetaldehyde] T refers to the sum of the equilibrium concentrations of the three major forms of acetaldehyde in model wine solution (free, hydrate, and ethyl hemiacetal). The constants for glycerol acetals were also calculated with the sum of the four isomers formed. Glycerol Acetals [total glycerol acetals] Keq = [acetaldehyde] T [glycerol] (1) ethylidene-bridged Catechin K eq = [ethylidene bridged catechin dimers] 2 [acetaldehyde] [( + ) catechin] T Glutathione Adducts [GSH adducts] Keq = [acetaldehyde] T [glutathione] (3) RESULTS AND DISCUSSION A method for investigating acetaldehyde interactions and their resulting equilibria by 1D and 2D 1 H NMR was developed. The presence of 12% ethanol in model solutions limited the sensitivity of 1D 1 H NMR to low-concentration substances as the intensities of the ethanol signals obscured some other signals. A significantly increased sensitivity was accomplished by using deuterated ethanol (d 6 -EtOD), which gives decreased signals. A pulse sequence with presaturation further improved sensitivity by suppression of the large water signal at 4.69 ppm. The relaxation delay was set to 5.0 s for 1D experiments to ensure all molecules were relaxed prior to acquisition. Freeze thaw degassing and flame sealing of reaction solutions in NMR tubes under vacuum ensured the removal and exclusion of any oxygen over the course of the reaction. The sealed, oxygen-free system also ensured that acetaldehyde could not volatilize out of the tube, although a small amount was likely in the vapor form within the tube. The optimized method was used to determine the equilibrium constants for the acetalization reaction of glycerol, the formation of ethylidene bridges between flavan-3-ol units (specifically (+)-catechin units), and the thiol addition reaction with glutathione in model wine solutions. The ratios of reactant concentrations for each interaction were chosen to reflect those found in the typical wine matrix. The acetalization reaction of glycerol was performed with glycerol in 10 and 20 times molar excess, as this would provide the pseudo-first-order kinetics that are likely to occur in real wine. The reaction between catechin and acetaldehyde was run at 1:1 and 2:1 ratios ([catechin]: [acetaldehyde] T )toreflect the stoichiometry of reactants in the intermediate and bridged dimer product. The glutathione to acetaldehyde ratios studied were 1:1 and 10:1 to reflect low and excess levels of glutathione, respectively. Both ratios are possible in a wine medium, depending on the glutathione additions made. Assignments of Free Acetaldehyde, Hydrate, Ethyl Hemiacetal, and Diethyl Acetal. Within a simple model wine solution, there are two alcohol nucleophiles that can engage in addition reactions with acetaldehyde: water and (2) 6871

ethanol. A 1 H NMR spectrum of a simple model wine solution (12% ethanol, 6 g/l tartaric acid, ph 3.5) containing acetaldehyde confirms the presence of three major forms of acetaldehyde: the free carbonyl, its hydrated form 4, and ethyl hemiacetal 3a from the addition of one ethanol molecule to the carbonyl carbon (Figure 2). The expected signals for free Figure 2. 1 H NMR spectra of model wine solution spiked with diethyl acetal 3b. Over the course of a few hours, the full acetal hydrolyzes to its hemiacetal 3a, free acetaldehyde, and acetaldehyde hydrate 4. acetaldehyde are present with a doublet at 2.22 ppm for the methyl protons and a quartet at 9.65 ppm for the carbonyl proton. The hydrated form is signified by a doublet at 1.31 ppm and a quartet at 5.23 ppm. Finally, a doublet at 1.29 ppm and a quartet at 4.93 ppm were present, corresponding to the aldehydic methyl and methine protons of 3a (Table 1). All signal couplings were acquired with a COSY spectrum of the solution (Figure 3, top) and agree with previous reports. 29 Table 1. Assignment of 1D 1 H NMR Signals for Quantitation and Correlated 2D 1 H 1 H NMR Signals for Acetaldehyde Derivatives in Model Wine Solutions compound 1 H signal used for quantification (ppm) 2D 1 H 1 H correlation assignment acetaldehyde 2.22, d, 2.95 Hz 9.65 CH 3 glycerol 3.75, m 3.63, 3.54 CH 1a 1.35, d, 4.89 Hz 5.16 CH 3 1b 1.38, d, 4.89 Hz 5.07 CH 3 2a 4.12, dd, 12.62 Hz, 1.46 Hz 3.83, 3.47 H2 a 2b 3.94, dd, 10.60 Hz, 5.21 Hz 4.03, 3.69 H2 a 3a 1.29, d, 5.29 Hz 4.93 CH 3 4 1.31, d, 5.29 Hz 5.23 CH 3 a Assignments for 2a and 2b are based on the numbered structure in Figure 3. The identity of 3a was determined by the addition of diethyl acetal 3b standard to model wine. A spectrum taken immediately displayed a doublet at 1.29 ppm in addition to the two doublets adjacent that were previously described. In the COSY spectrum, this doublet was found to be coupled to a quartet at 4.78 ppm. Within a few hours, this signal area decreased by about 90% (hydrolysis of the acetal), whereas the signal areas for acetaldehyde, the hydrate, and the hemiacetal all increased by this amount distributed according to their relative ratios to each other (Figure 2). Analysis of Acetaldehyde Equilibria in Model Wine. Throughout the course of reactions subsequently investigated, the ratio between the free acetaldehyde, acetaldehyde hydrate, Figure 3. 1 H 1 H 2D COSY spectra of the equilibrium between glycerol, acetaldehyde, hydrate 4, ethyl hemiacetal 3a, and the resulting four glycerol acetal isomers 1a,b and 2a,b: (top) coupled methyl and methine proton signals of 1a,b, 2a,b, 3a, and 4; (bottom) crosscouplings for alternate signals used for quantitation of 2a and 2b (H2) due to obstruction of methyl signals for 2a,b by those of 3a and 4 in 1D spectra. and ethyl hemiacetal remained at 1:1.15:0.828. The full diethyl acetal was also present in all reaction solutions, although its abundance was very low relative to the hemiacetal. Equilibrium constants for the basic forms of acetaldehyde in model wine were calculated and are reported in Table 2. Previously Table 2. Equilibrium Expressions and Resulting Equilibrium Constants for the Hydrate, Hemiacetal, and Ethyl Acetal Forms of Acetaldehyde in a Model Wine Solution at ph 3.5 and 30 C compound acetaldehyde hydrate 4 ethyl hemiacetal 3a diethyl acetal 3b equilibrium expression [hydrate] [free acetaldehyde] [hemiacetal] [free acetaldehyde][etoh] [diethyl acetal] [free acetaldehyde][etoh] 2 calcd equilibrium constant 1.15 ± 0.04 0.412 ± 0.009 M 1 0.00880 ± 0.00015 M 2 published equilibrium values for the hydrate (1.06), 9 ethyl hemiacetal (0.496), 39 and ethyl acetal (0.0125) 39 are in good agreement with calculated constants. The immediate distribution of acetaldehyde into 4, 3a, 3b, and the carbonyl form indicates that although the content of all three contribute to the investigated interactions, only the amount in the free form imparts an aldehydic aroma to oxidized wines. Studies of sherry and other commonly fortified and aged wines have reported 3b 6872

to give off a licorice/green fruit aroma and to contribute to the overall fruity and balsamic aroma of these wines. 10 The hemiacetal was previously assumed to be a fleeting intermediate in alcoholic beverages, most likely because it is too unstable to be measurable by conventional methods such as GC, 40 during which the heat of vaporization would cause dissociation. The fact that such a large fraction of acetaldehyde is present in the ethyl hemiacetal form in wine and model wine solutions has only come to light with NMR analysis of acetaldehyde in wine. 29 Through NMR analysis, we have determined that only a portion of the total acetaldehyde present is in its free and active form under wine-relevant conditions, but 4 and 3a provide reserves of acetaldehyde that immediately feed into the free form as it is depleted through other reactions. Because all three forms of acetaldehyde contribute to the formation of the product in this way, the combined concentration of free, hydrated, and acetalized acetaldehyde ([acetaldehyde] T in eqs 1 3) at equilibrium is used for the determination of apparent equilibrium constants. Assignments of Glycerol Acetal Isomers. The proton resonances for glycerol were confirmed in a spectrum of the compound in the model wine solution. Two equivalent pairs of protons give two doublet of doublets at 3.63 and 3.54 ppm. The methine proton on the center carbon is split by both pairs of protons, giving a multiplet signal at 3.75 ppm; the latter was used for quantitation of glycerol. Proton resonances for the glycerol acetal isomers 1a,b and 2a,b were established with 1D and 2D 1 H NMR analysis of the synthesized standard (containing all four isomers) in tartrate buffer at ph 3.5. Signals for the protons in each isomer were established by first grouping them by COSY cross-couplings and then assigning them to each isomer according to structure, number of signals, and relative abundance. The dioxolane isomers (1a,b) were predicted to have five distinct signals downfield as all five protons in the molecule are unique. The dioxane structure (2a,b) lends itself to only three signals because there are two pairs of protons that are equivalent and thus share the same signal (Supplementary Figure 1). NMR signal assignments for all protons of the glycerol acetal isomers can be found in Tables 1a and 1b of the Supporting Information. The signals with the highest abundance in the 1D spectra were attributed to the methyl protons from acetaldehyde. Four doublets at 1.28, 1.29, 1.35, and 1.38 ppm were assigned to trans-dioxane, cis-dioxane, trans-dioxolane, and cis-dioxolane, respectively, on the basis of prior knowledge of the relative amounts of the isomers in the standard. 41 These doublet signals were correlated with four quartets at 4.78, 4.88, 5.16, and 5.07 ppm, representing the methine protons that are within two bonds of the methyl protons (Figure 3). Being the most abundant signals, the methyl proton resonances for cisdioxolane 1b and trans-dioxolane 1a were used for quantitation. Due to the obstruction of their methyl proton signals by those of 4 and 3a, cis- and trans-dioxane isomers had to be quantitated by two signals further downfield (Figure 4). cis- Dioxane 2b was quantitated by a doublet of doublets at 3.94 ppm with COSY coupled signals at 4.03 and 3.69, whereas trans-dioxane 2a was quantitated by a doublet of doublets at 4.12 ppm with COSY coupled signals at 3.83 and 3.47 ppm (Figure 3; Table 1). Analysis of Equilibria of Glycerol Acetal Reactions. At the median ph (3.5) and at 40 C, dioxolanes 1a and 1b were found to reach equilibrium after about 280 h, whereas dioxanes Figure 4. 1 H NMR spectrum of the equilibrium between glycerol, acetaldehyde, hydrate (4), hemiacetal (3a), and the resulting dioxolane (1ab) and dioxane (2a,b) isomers. The signal for d 4 -TMSP was calibrated to 0 ppm. 2a and 2b took more than twice this time to equilibrate. Dioxolane isomers could be detected within 24 h of reaction under the conditions specified above. The partially obstructed methyl signals were detected after approximately 60 h of reaction, but still could not be used for quantitation at equilibrium. As shown in Table 3, equilibrium constants were Table 3. Apparent Equilibrium Constants and Acetaldehyde Consumed for the Reaction between Glycerol and Acetaldehyde in Model Wine Solutions at ph 3 4 and 25 50 C a [glycerol]: [acetaldehyde] ph, temperature ( C) apparent equilibrium constant b (M 1 ) [acetaldehyde] consumed (%) c (%) 10:1 3.0, 40 2.53 ± 0.043 20.2 10:1 3.5, 25 0.959 ± 0.052 10.6 10:1 3.5, 40 1.64 ± 0.030 16.1 20:1 3.5, 50 1.73 ± 0.006 33.5 10:1 3.4, 50 1.71 ± 0.11 19.1 10:1 4.0, 40 1.14 ± 0.056 12.2 a Average acetaldehyde concentration was 8.50 mm. b Calculated using the sum of the concentrations of all four acetal isomers for [acetals] in eq 1. c Percentage calculated by dividing total acetal content by starting total acetaldehyde concentration. slightly greater than unity, and the percent acetaldehyde consumed was 16% at ph 3.5 when the glycerol concentration was 10 times the molar concentration of acetaldehyde. With a major portion of the initial acetaldehyde content remaining at equilibrium, the formation of acetals from glycerol is not a highly favorable reaction in wine. This was expected on the basis of the formation of water in the acetal reaction and the high concentration (49 M) of water present under wine conditions. Equilibrium constants for the six-membered ring isomers and the five-membered ring isomers were also calculated separately (Figure 5a). At the median ph (3.5), the dioxane versus the dioxolane contents present at equilibrium were relatively similar (despite differences in time to equilibration), indicating that at ph 3.5, neither isomer group is favored over the other at equilibrium. At ph 3.0, however, the equilibrium constant ratio reveals that the dioxane isomers are more highly favored than the dioxolanes. The K dioxanes is almost doubled when the reaction is run in a lower ph environment, whereas K dioxolanes remains just below unity. At ph 4.0, the formation of all acetal 6873

Table 4. Assignment of 1D 1 H NMR Signals for Quantitation and Correlated 2D 1 H 1 H NMR Signals for (+)-Catechin and Ethylidene-Bridged (+)-Catechin Dimer Isomers in Model Wine solutions compound 1 H signal used for quantification (ppm) 2D 1 H 1 H correlation assignment a (+)-catechin 2.54, dd 2.86, 4.21 H-4α 5a 1.48, d, 7.69 Hz 4.90 CH 3 -bridge 5b 1.59, d, 7.69 Hz 4.79 5c 1.72, d, 7.53 Hz 4.82 a Proton assignments are based on numbered structures in Figure 7. Figure 5. Effects of (a) ph and (b) temperature on equilibrium constants for the formation of dioxolanes versus dioxanes. Calculated equilibrium constants are also reported in Table 2 of the Supporting Information. isomers is decreased, but again K dioxanes is more drastically affected. The acetalization of glycerol is an acid-catalyzed reaction, so the rate of reaction was expected to be affected by the change in ph, and previous studies of acyclic acetal formation in alcoholic solutions showed that equivalent concentrations of the acetal were formed at various ph values with only the rate being affected. 40 Although this result was surprising, our data clearly show that ph alters the equilibrium constants for the formation of cyclic acetals. As shown in Figure 5b, an increase in reaction temperature resulted in greater formation of all acetal isomers, indicating an endothermic reaction. The change in temperature caused the most significant increase in the dioxane isomer content, whereas the dioxolane isomer content was only slightly altered. As the thermodynamic products of the glycerol acetal reaction, the dioxane isomers were expected to form more slowly than the dioxolane isomers, 42 but dominate the total acetal content at higher temperatures. This was confirmed by the results of the change in reaction temperature. Assignments of Ethylidene-Bridged Catechin. The proton resonances for ethylidene-bridged catechin dimers 5 were established by analysis of model wine solutions containing catechin and acetaldehyde over time (Table 4) and were assigned according to the numbered dimer structure in Supplementary Figure 2. Signals for (+)-catechin alone were confirmed with the literature. 43 The proton H-4α was assigned to the doublet of doublets at 2.54 ppm and was used for quantitation of (+)-catechin molecules. Following the addition of acetaldehyde, three doublet signals at 1.48, 1.59, and 1.72 ppm gradually increased over time (Figure 6). These signals were attributed to the methyl protons on the bridge between two catechin units. The presence of two positions on the A ring (C6 and C8) that can be involved in an electrophilic substitution reaction results in three possible acetaldehyde bridge connections between two catechin units: 6 6, 6 8, and 8 8. The presence Figure 6. 1 H NMR spectrum of the equilibrium established between bridged catechin dimer isomers 5a c, (+)-catechin, free acetaldehyde, ethyl hemiacetal 3a, and the hydrate 4. In the tartrate buffer without ethanol, isomer b is resolved into two overlapping and equivalent doublets. The assigned structures of 5a, 5b, and 5c are given to the right of the spectrum. of three doublets is representative of these three connections and in agreement with previously reported data for the formation of bridged catechin dimers. 27 In the 6 8 isomer, the bridging carbon is a chiral center, yielding two diastereomers for this dimer. The formation of both is apparent when the catechin acetaldehyde reaction is investigated in aqueous solutions without ethanol. This slight change in chemical environment increases the resolution between two nearly identical bridge methyl signals. The overlapping doublets are thus presumed to be the 6 8 stereoisomers (5b, Figure 6). Drinkine et al. reported the 8 8 dimer isomer to have formed in the most abundance, whereas the 6 6 isomer was formed in very small quantities. 27 On the basis of these data, we deduce that 5a is the 8 8 isomer (highest yields), whereas 5c is the 6 6 isomer (Figure 6), but these assignments cannot be confirmed without further assessment. Coupled signals representing the methine protons of the ethylidene bridges were determined (by 2D COSY NMR) to be 4.90, 4.79, and 4.82 ppm for a, b, and c, respectively (Figure 7). The signals we report here are in good agreement with the signal ranges reported for ethylidene-bridged catechin polymers by Kim et al. in organic solvent. 44 Signals reported by Saucier et al. for 5a 5c are slightly different due to the disparity between solvents in which the isomers were analyzed. 45 For each unique proton resonance present for a single catechin molecule, three new sets of signals with similar chemical shifts appeared near catechin signals in 1D and 2D spectra of the equilibrated reaction solution. These adjacent signals represent the protons of two catechin molecules bridged by condensation with acetaldehyde (enlargement of signals around 2.45 ppm, Figure 6). These 6874

acid content to hydrolyze stable dimers (Table 5). The acetaldehyde consumed through this reaction was similar to Table 5. Calculated Equilibrium Constants and Acetaldehyde Consumed for the Reaction between Acetaldehyde and (+)-Catechin in Model Wine Solutions at ph 3 4 and 40 C a Figure 7. 1 H 1 H COSY spectrum of coupled signals for the ethylidene bridge between two catechin units. The three cross peaks represent isomers 5a, 5b, and 5c. Many of the surrounding cross peaks are unidentifiable because they correlate to low-abundance signals in 1D spectra, but could be intermediates or side-reaction products. correlations are detailed further in Table 3 of the Supporting Information. Throughout most of the reaction, signals of very low intensity can be detected by 2D COSY at 1.46 ppm and correlated to 5.27 and 5.37 ppm (Figure 7). It is possible that these signals represent the methyl protons of the catechin ethyl adduct, the intermediate that forms before a second catechin unit has reacted (Supplementary Figure 3). Predicted NMR signals for the intermediate structures indicate the methine proton from acetaldehyde would be shifted further downfield from the signal for the bridge methine proton in the full dimer. Additional signals of low intensities appeared in spectra from 1.40 to 1.50 ppm with 2D correlation signals around 3.7 and 4.65 ppm. Some of these signals may represent bridges formed between dimers and a third catechin unit. Because potential trimer bridge signals did not overlap with dimer bridge signals, their intensities in the first-dimension spectra were very low and could not be quantitated accurately. Analysis of Equilibrium of Acetaldehyde Condensation of Catechin. The formation of ethylidene-bridged (+)-catechin dimer isomers 5a 5c from equimolar concentrations of acetaldehyde and catechin at ph 3.5 reached equilibrium in approximately 200 h. After this point, spectra indicated that some portion of the dimers continued to react with acetaldehyde, most likely to form dimer ethyl adducts and ethylidene-bridged trimers. Kim et al. notes that the dimer and trimer are the major products from polymerization of equimolar amounts of (+)-catechin and acetaldehyde. 44 Eventually, a haze developed in reaction solutions with some precipitation occurring as well. Previous studies of this reaction have reported haze and precipitation formation when acetaldehyde is in molar excess to catechin, conditions under which bridged trimers and tetramers form and aggregate. 44,46 Because reactions were studied with equimolar concentrations of reactants, precipitation was not initially expected. Spectral evidence indicates that some trimer formation likely occurred in reaction solutions, but that the bridged dimers were the major products of this interaction. Thus, their aggregation and oversaturation of the reaction solution are believed to be the cause of precipitation in reaction solutions. Equilibrium constants for the formation of 5a 5c indicate a higher thermodynamic favorability compared to acetals, which was expected on the basis of reports of the need for increased 6875 [catechin]:[acetaldehyde] concentration ratio ph apparent equilibrium constant b (M 2 ) [acetaldehyde] consumed (%) 1:1 3.0 (6.13 ± 0.32) 10 3 16.5 2:1 3.5 (5.00 ± 0.15) 10 3 16.2 1:1 3.5 (3.92 ± 0.13) 10 3 13.0 1:1 4.0 (5.40 ± 0.59) 10 3 12.8 a Average concentration of acetaldehyde was 8.45 mm for equimolar reactions and 3.85 mm for the 2:1 reaction. b Content of all ethylidenebridged catechin dimer isomers was summed and used in the equilibrium constant calculation. that determined for the glycerol acetal reaction, although the nucleophile (catechin) was not in molar excess to acetaldehyde. The product concentration in the numerator of the expression did not include all products formed because it was not possible to quantitate intermediates or possible trimers as the signal abundances in the 1D spectra were too low. Unfortunately, more information on the formation of the trimer could not be obtained from hazy solutions given that the inhomogeneity results in poor line shape of spectra and inaccurate quantitation. This is an obvious limitation of the application of NMR to the analysis of bridged catechin dimer formation. However, NMR has enabled the quantitation of bridged dimers yielded from equimolar concentrations of acetaldehyde and (+)-catechin, which has not been reported previously. The variability of the calculated equilibrium constants was higher than that of the constants for acetals and GSH adducts. The complexity of the molecules involved causes more variable quantitation and is compounded in the denominator of the equilibrium expression with the square of the catechin concentration (eq 2). The different ph values at which the bridging reactions were observed did not alter the overall consumption of acetaldehyde toward formation of 5a 5c, but it appeared to affect the rates of the reactions investigated (to be analyzed in future work). In general, similar concentrations of the dimers resulted from the reaction between equimolar amounts of (+)-catechin and acetaldehyde at differing ph levels. Thus, the variation in equilibrium constants between different ph levels is due to the reactant concentrations. Total losses in reactants could not be fully accounted for with dimer quantitation as other products formed after their equilibration that could not be quantitated. In reaction solutions at ph 4.0, there was a higher loss of catechin and acetaldehyde despite a lower content of 5a 5c. This unexpected outcome may be due to the increased time required to reach equilibration, potentially yielding various catechin polymers that were likely below the limit of quantitation. Assignments of Glutathione Thiohemiacetal. The proton resonances for the glutathione thiohemiacetal 8 were determined by 1D and 2D NMR analysis of model wine solutions containing GSH alone and with acetaldehyde (Table 6). Chemical shifts for GSH were confirmed by comparison with those reported (in D 2 O) in the literature. 47 Sonni et al. reported a doublet at 1.36 ppm in a tartrate buffer in D 2 O, 20 representing the GSH adduct 8. Analysis of solutions

Table 6. Assignment of 1D 1 H NMR Signals for Quantitation and Correlated 2D 1 H 1 H NMR Signals for Glutathione and Thiohemiacetals in Model Wine Solutions compound 1 H signal used for quantification (ppm) 2D 1 H 1 H correlation assignment glutathione 3.80, t 2.19 CH 8 1.49, d, 6.49 Hz 5.11 CH 3 8a 1.68, d, 6.48 Hz 4.96 CH 3 1.70, d, 6.48 Hz 4.91 containing both acetaldehyde and GSH by H NMR in model wine revealed two overlapping doublets at 1.48 ppm. These signals were attributed to the methyl protons of acetaldehyde in the thiohemiacetal formed. The presence of two doublets with similar abundance and identical 2D coupling indicated that two diastereoisomers of 8 form to the same degree, where the carbonyl carbon bound to the glutathione sulfur is the chiral center. A 2D COSY spectrum gives the correlated signal at 5.11 ppm, representing the methine proton bound to the chiral carbon (Figure 8). The overlapping doublets for the methyl protons of 8 were integrated for its quantification. NMR spectra of reaction solutions containing glutathione in molar excess to acetaldehyde (10:1) displayed an additional pair of overlapping doublets (of lower abundance) further downfield at 1.68 and 1.70 ppm. 2D NMR analysis of these signals gave COSY correlations at 4.96 and 4.91 ppm, respectively. Investigations into the reaction between GSH and formaldehyde yielded evidence of the cyclization of the aldehyde-bound portion of GSH when glutathione was in excess and acid was present. 48 Supplementary Figure 4 displays the predicted structure for the second product (8a) formed from the reaction. This structure contains a chiral center at the same carbon as in 8, which would give rise to two diastereoisomers and thus two doublets with very similar chemical shifts. The secondary product only accounted for an average of 4.31% of the total content of GSH adducts and is not a major path by which acetaldehyde is consumed. Analysis of Equilibrium of Glutathione Addition to Acetaldehyde. The addition of the thiol portion of glutathione to the carbonyl carbon of acetaldehyde was a very rapid reaction, reaching equilibrium within a few minutes. This is similar to the rapid addition of bisulfite to acetaldehyde. However, unlike the complete consumption of acetaldehyde by an equimolar amount of bisulfite, an equimolar concentration of glutathione consumes <10% of acetaldehyde present. The reciprocals of the equilibrium constants displayed in Table 7 Table 7. Calculated Equilibrium Constants and Acetaldehyde Consumed for the Reaction between Acetaldehyde and Glutathione in Model Wine Solutions at ph 3.0 4.0 and 30 C a glutathione/acetaldehyde concentration ratio ph calcd equilibrium constant (M 1 ) acetaldehyde consumed (%) 1:1 3.0 10.35 ± 0.10 9.56 10:1 3.5 11.02 ± 0.32 52.4 1:1 3.5 11.17 ± 0.47 9.78 1:1 4.0 10.18 ± 0.22 9.79 a Average concentration of acetaldehyde was 8.34 mm. Figure 8. NMR analysis of GSH adducts 8 and 8a resulting from interaction between acetaldehyde (including its hydrate 4 and ethyl hemiacetal 3a) and glutathione: (top) 1D spectrum, reaction solution containing equimolar amounts of the two reactants and the resulting stereoisomers 8, with partial structure shown; (bottom) 2D COSY spectrum, reaction solution containing GSH in excess, producing 8 and an additional pair of stereoisomers 8a. indicate the dissociability of the adduct and can be compared with dissociation constants reported for both bisulfite and glutathione addition products of acetaldehyde. The dissociation constant reported for hydroxyethylsulfonic acid 7 (2.06 10 6 ) 49 is much smaller than this value calculated for the GSH adduct because acetaldehyde is less tightly bound by glutathione than it is by bisulfite. Kanchuger et al. 50 reported the observed equilibrium constant (K obsd ) for the thiohemiacetal 8 as 12 ± 1M 1 in a ph 4 buffered solution, whereas Sonni et al. 20 determined the dissociation constant for the adduct in model wine to be (5 ± 1) 10 2 M at ph 3.0. Our values agree well with those reported by Sonni et al. 20 when the concentration of the carbonyl form. This is almost twice the dissociation constant determined here and by Kanchuger et al. 50 for this interaction due to the inclusion of the hydrate and ethyl hemiacetal forms in the acetaldehyde concentration used for calculations. Our values agree well with Sonni et al. 20 when the concentration of the carbonyl form of acetaldehyde alone is used to calculate the constant. The ph had little effect on the equilibrium constants as the thiol reaction is an not an acidcatalyzed reaction. The reaction has been reported to be independent of ph if the aldehyde involved lacks an ionizing moiety. 50 Implications of Acetaldehyde Reaction Equilibrium Constants. Through the measurement of acetaldehyde reaction solutions by 1D and 2D NMR, equilibrium constants were measured for the acetalization of glycerol, condensation of 6876

(+)-catechin, and formation of a thiohemiacetal of glutathione. In addition, this method revealed the major forms of acetaldehyde within a wine-like medium: the free carbonyl, the hydrate, and ethyl hemiacetal. The equilibrium between these forms is labile, all forms adjusting as the free aldehyde is consumed. The glycerol acetals were the least thermodynamically favored, which was expected due to the formation of water in the reaction and its high concentration in the reaction solution. Glutathione and (+)-catechin consumed acetaldehyde to a similar degree in reactions with equimolar concentrations of reactants. Unlike SO 2, even a large molar excess of glutathione did not completely consume acetaldehyde to form a thiohemiacetal. On the basis of the equilibrium constants calculated for the bridging of (+)-catechin molecules, the flavanol dimers appear to be the most thermodynamically favored and may be stable enough to be measured and employed as an oxidation marker over the short term. Using a method that can measure one or more products of acetaldehyde without altering the equilibria, equilibrium constants can then be used to estimate the level of free acetaldehyde present. In all equilibria investigated in this study, a significant fraction of the initial acetaldehyde content remained unreacted, implying that acetaldehyde formed in oxygenated wines is not completely consumed by any specific nucleophile (except free SO 2 ). The extent to which acetaldehyde is consumed by each nucleophile studied provides implications for the consumption of acetaldehyde in real wines. To understand the competition between these nucleophiles and determine the fate of acetaldehyde in the presence of multiple nucleophiles, the kinetic properties of the reaction studied must also be determined. This will be investigated in the future through NMR measurements taken throughout the course of the reactions. It is possible that this method could be applied to characterize other important acetaldehyde interactions that occur in wine. Comparison between the thermodynamic and kinetic properties of various acetaldehyde reactions and a specific wine s composition will make it possible to predict the expected reactions of the acetaldehyde from oxidation and the persistence of those products. Despite the limitations of NMR for the analysis of many these acetaldehyde reactions products in real wine, its utilization in this work as a tool to explore acetaldehyde chemistry in wine will provide a means of estimating acetaldehyde levels from the concentration of a more stable reaction product measured by other methods. Equilibrium constants determined in this study will enable better predictions of acetaldehyde consumption in a wine based on its composition, the determination of the cumulative amount of acetaldehyde in all its forms in a wine, and, ultimately, an understanding of the fate of acetaldehyde produced through oxidation of a specific wine. This investigation has characterized the major reversible reactions of acetaldehyde, but additional reactions, such as the formation of vitisin B, will be necessary for a complete picture. Considering the widespread use of micro-oxygenation to accelerate aging in red wines and the importance of controlled oxidation in bottle aging, developing a means to accurately observe and predict oxidation products will provide new tools to properly manage the process. 6877 ASSOCIATED CONTENT *S Supporting Information The Supporting Information is available free of charge on the ACS Publications website at. 2D 1 H 1 H COSY NMR assignments for cis- and transdioxolane and dioxane isomers (Table 1), calculated equilibrium constants for dioxane and dioxolane isomers (Table 2), and ethylidene-bridged dimers (Table 3); figures showing proton assignments for acetaldehyde addition products (PDF) AUTHOR INFORMATION Corresponding Author *(A.L.W.) E-mail: alwaterhouse@ucdavis.edu. Phone: (530) 752-4777. Notes The authors declare no competing financial interest. ACKNOWLEDGMENTS We thank Bennet Addison of the Nuclear Magenetic Resonance Facility at the University of California, Davis, for providing assistance with and advice on NMR analyses. We acknowledge Roger B. Boulton for his assistance through discussions about reaction equilibria. The American Vineyard Foundation provided financial support. ABBREVIATIONS USED NMR, nuclear magnetic resonance; GSH, glutathione; SO 2, sulfur dioxide; d 6 -EtOD, deuterated ethanol-d 6 ; d 4 -TMSP, deuterated 3-trimethylsilylpropionic-2,2,3,3-d 4 acid sodium salt; COSY, correlation spectroscopy; 1D 1 H NMR, onedimensional proton nuclear magnetic resonance; 2D 1 H 1 H NMR, two-dimensional homonuclear proton correlation spectroscopy nuclear magnetic resonance REFERENCES (1) Pasteur, M. L. Etudes sur le Vin, 2nd ed.; Librairie F. Savy: Paris, 1875; p 544. (2) Danilewicz, J. C. Interaction of sulfur dioxide, polyphenols, and oxygen in a wine-model system: central role of iron and copper. Am. J. Enol. Vitic. 2007, 58, 53 60. (3) Elias, R. J.; Waterhouse, A. L. Controlling the Fenton reaction in wine. J. Agric. Food Chem. 2010, 58, 1699 1707. (4) Oliveira, C. M.; Ferreira, A. C. S.; De Freitas, V.; Silva, A. M. S. Oxidation mechanisms occurring in wines. Food Res. Int. 2011, 44, 1115 1126. (5) Carlton, W. K.; Gump, B.; Fugelsang, K.; Hasson, A. S. Monitoring acetaldehyde concentrations during micro-oxygenation of red wine by headspace solid-phase microextraction with on-fiber derivatization. J. Agric. Food Chem. 2007, 55, 5620 5625. (6) da Silva Ferreira, A. C.; Barbe, J.-C.; Bertrand, A. Heterocyclic acetals from glycerol and acetaldehyde in Port wines: evolution with aging. J. Agric. Food Chem. 2002, 50, 2560 2564. (7) Drinkine, J.; Lopes, P.; Kennedy, J. A.; Teissedre, P.-L.; Saucier, C. Analysis of ethylidene-bridged flavan-3-ols in wine. J. Agric. Food Chem. 2007, 55, 1109 1116. (8) Ahrens, M.-L.; Strehlow, H. Acid catalyzed hydration of acetaldehyde. Discuss. Faraday Soc. 1965, 39, 112 120. (9) Kurz, J. L. Hydration of acetaldehyde. I. Equilibrium thermodynamic parameters. J. Am. Chem. Soc. 1967, 89, 3524 3528. (10) Moyano, L.; Zea, L.; Moreno, J.; Medina, M. Analytical study of aromatic series in sherry wines subjected to biological aging. J. Agric. Food Chem. 2002, 50, 7356 7361.