INTERACTIONS BETWEEN ANTHOCYANINS, PHENOLIC COMPOUNDS, AND ACETALDEHYDE AND THEIR SIGNIFICANCE IN RED WINES

INTERACTIONS BETWEEN ANTHOCYANINS, PHENOLIC COMPOUNDS, AND ACETALDEHYDE AND THEIR SIGNIFICANCE IN RED WINES C. F. Timberlake and P. Bridle Cider and Fruit Juices Section, Long Ashton Research Station, Bristol, BS18 9AF, England. Thanks are due Dr. G. Hrazdina, New York State Agricultural Experiment Station, Geneva, New York, for the gift of some Of the anthocyanins; Dr. L. Jurd, Western Regional Laboratory, ARS, Berkeley, California, for a gift of 7-hydroxy-6-methyl flavylium chloride; and our colleagues Dr. F. W. Beech and Mr. A. G. H. Lea for helpful discussion of some aspects. Accepted for publication July 21, 1976. ABSTRACT Interactions between pure anthocyanins [malvi- with extent of the shift varying with the type of din (Mv) 3-glucoside, Mv 3,5-diglucoside, and Mv component. The orders of decreasing reactivity 3-p-coumarylglucoside-5-glucoside], phenolic com- were Mv 3-glucoside > Mv 3-p-coumarylglucosidepounds [ (-)-epicatechin, (+)-catechin, (-)-epigal- 5-glucoside > Mv 3,5-diglucoside, and trimer C1 locatechin, procyanidin dimer B2, and procyanidin = dimer B2 > epicatechin. Conversely the extrimer C1], and acetaldehyde were studied in tar- tent of color increase and its stability decreased trate buffer, ph 3.5, at room temperature in air in reverse order, with Mv diglucoside > acyunder darkness and light. Several reactions occurred lated diglucoside > monoglucoside, because of simultaneously. The anthocyanins and phenolic precipitation with the latter. Color augmentation compounds reacted very slowly, with eventual formation of yellow xanthylium salts (one presumed glycosylated, others sugar-free), confirming that condensation occurred via the anthocyanin 4-position. Unreacted procyanidins B2 and C1 were partly transformed to epicatechin and lower and higher polymeric procyanidins, the reactions being catalyzed by daylight. There was little reaction between the anthocyanins and acetaldehyde except with Mv 3-glucoside, which was slowly polymerized. However, the addition of acetaldehyde to mixtures of anthocyanins and phenolics caused rapid and spectacular color augmentation with shifts toward violet, was due to the formation of highly colored new compounds, detectable by chromatography and believed to consist of anthocyanins and phenolics linked by CH~CH bridges. A reaction scheme is proposed in which the initial reaction product of acetaldehyde and phenolic combines with the anthocyanin at position 8, followed by anhydrobase formation. A small amount of a compound resulting from interaction of Mv 3,5-diglucoside, (+)-catechin, and acetaldehyde was isolated and its properties examined. The significance of the findings are discussed in relation to color changes of red wines. Once anthocyanins leave the environment of the fruit cell in which they are stabilized, they can undergo a series of reactions leading to their structural modification or degradation (10). Thus, during the making of red wine the monomeric anthocyanins of the grapes are converted into reddishbrown polymeric pigments (23,25) by their reaction with other phenolic components, probably procyani- 97 dins (24). Synthetic flavylium salts, structurally related to the anthocyanins, are known to condense with certain phenolic compounds at position 4 (Fig. 6) of the flavylium structure, forming discrete identified products (13,14,15), although similar reactions have not yet been demonstrated with anthocyanins themselves. Studies on bisulfite resistance (29) and ionization characteristics (24)

98mlNTERACTIONS BETWEEN ANTHOCYANINS, PHENOLIC COMPOUNDS, AND ACETALDEHYDE suggest that condensation of anthocyanins occurs likewise via position 4, although the only direct evidence appears to be the finding of a glucosylated xanthylium salt in stored grape juice (11). Very little work has been reported on interactions between pure anthocyanins and specific phenolics in isolated systems. Somers (24) examined changes in a model wine extract from grape skins but, of necessity, without any control solution, and Rib- 6reau-Gayon (21) concluded that some interaction occurred between malvidin 3,5-diglucoside and pinetannin, with and without iron salts. Thus there seemed need for a study of simple systems of wine ph containing known pure components with appropriate control solutions carried out at room temperature rather than elevated temperatures, which might accelerate changes (12) but also risk altering the reaction course. Any study should include anthocyanins of several types having somewhat different properties and, among the phenolics, especially procyanidins akin to those present in wines.~ viz., the dimers B1 (31), B2, B3, and B4 (8). Additionally, interactions with acetaldehyde should be included because of its reported effect during wine aging (24,30), as well as a study of other factors (iron, light, ph). Accordingly, interactions were studied between anthocyanins representative of three different types [malvidin (Mv) 3-glucoside, Mv 3,5-diglucoside, and Mv 3-p-coumarylglucoside-5-glucoside] and phenolic compounds of differing molecular size and complexity [the monomeric catechins, (-)-epicatechin, ( + )- catechin, and (-)-epigallocatechin, procyanidin dimer B2, and procyanidin trimer C1], with and without iron salts or acetaldehyde, in darkness and in daylight, at various ph values and with various buffer anions. Solutions were allowed to react spontaneously at room temperature in stoppered test tubes in the presence of air, and samples were withdrawn at intervals for examination of spectra and chemical changes in the components (by twodimensional paper and thin-layer chromatography). MATERIALS AND METHODS Preparation of solutions: The medium employed for most experiments was aqueous potassium hycirogen tartrate (0.02M; ph 3.5) containing ethanol (10% v/v). The concentrations of anthocyanins were usually 0.95 x 10-4M (about 0.5 mg/ml) ; the phenolic concentrations varied from 0.5 to 5 times those of the anthocyanins on a molar basis. In all experiments the acetaldehyde concentration was 0.2% v/v. Ferric chloride was added to give iron at 5 ppm. Phenolic compounds: (-)-epicatechin and (+)- catechin were commercial samples. Procyanidins B2 and C1 were isolated and characterized from cider (18,19). Both contain only (-) -epicatechin units. Chromatography: a) Paper: Small amounts of the reaction solutions were evaporated to dryness and dissolved in methanol (0.25 ml) for examina- tion. Ascending paper chromatography was carried out on Whatman No. 2 paper in sec-butanol-acetic acid-water 14"1"5; s-baw) as first dimension and acetic acid (2%; HOAc) as second. Additional anthocyanin solvents used were acetic acid~conc. HCl~water (15:3:82 v/v; HOAc-HC1) and n-butanol~acetic acid~water (6:1:2 v/v; BAW). Solvents used for xanthylium salts were solvent A, acetic acid~conc. HCl~water (20:5:80 v/v; solvent B, formic acid~3 N HC1 (1:1 v/v) and solvent C, acetic acid~conc. HCl~water (40:5:80 v/v). b) Thin-layer: Thin-layer chromatography (TLC) was on silica plates (self-indicating Merck $1 preparative in solvent toluene-acetone-formic acid (30:30:6 v/v) (6). Chromatograms containing phenolics spontaneously developed brown spots very soon after examination under UV light ((18). c) Column: Transformation of the procyanidins was examined by adsorbing samples onto small columns of Sephadex LH 20, eluting with ethanol, and identifying the products by paper and TLC. Spectra: Spectra were measured on a Unicam SP 800 recording spectrophotometer. Xanthylium salts were recorded first in 1% aqueous HC1 and then after: a) extraction into n-butanol (BuOH) ; and b) addition of alkali (NaOH) (11,17). Flavylium salts: 7-Hydroxy and 7-hydroxy-8-methyl flavylium chlorides were made by condensation of resorcylaldehyde or 2,4-dihydroxy-3-methylbenzaldehyde with acetophenone (15). RESULTS The results can be summarized as follows. Except where specified otherwise, the results refer to solutions of potassium hydrogen tartrate (0.02M) containing ethanol (10%; v/v) at ph 3.5, stored in darkness. Reactions of anthocyanins with phenolic compounds: It is known that anthocyanins become more colored by copigmentation with a wide range of flavonoids and other compounds (2) but that the copigment complexes are dissociated by addition of ethanol. Under our conditions (10% ethanol) some copigmentation occurred as evidenced by a significant and instantaneous color augmentation of the three anthocyanins on addition of the carechins, epigallocatechin, and the procyanidins. Augmentation increased with increasing concentrations of anthocyanin and phenolic and could amount to about 20%; it appeared greatest with malvidin (Mv) 3,5-diglucoside. On storage, mixtures of anthocyanins and phenolics gradually lost color in the red region (520 nm) but increased in the brown region (420.-450 nm). Similar behavior, however, was also shown by anthocyanins alone (which faded) and phenolics alone (which browned). It was thought that some evidence of interaction between the two might be obrained by comparing the spectra of mixtures of anthocyanins and phenolics with the sum of the spec-

!i HOAc oog ( G QO Fig. 1. Paper chromatography of components. X1 and X2, xanthylium salts (aglycones); X3, xanthylium salt (glycosylated); M, Mv 3-glucoside; ME, Mv 3-glucoside- acetaldehyde- epicatechin compound; D, Mv 3,5-diglucoside; D- cat, Mv 3,5-diglucoside - acetaldehyde- (+)-catechin compound; A, Mv 3-(pcoumarylglucoside)-5-glucoside; A- cat, Mv 3-(p-coumarylglucoside)-5-glucoside - acetaldehyde - (+)-catechin compound; E, (-)-epicatechin; cat, (+)-catechin; B2, dimeric procyanidin; C1, trimeric procyanidin; P, p-coumaric acids. tra of each component held separately. Because of the color augmentation referred to above, however, it was apparent that the latter spectra were not additive, and although a correction for augmenration could be made at the beginning of the experiment the same correction might not be valid after storage, when the composition of the system may have changed through interactions. The method is rather approximate, therefore, but within those limitations it nevertheless appeared that after seven months of storage a mixture of Mv 3-glucoside and epicatechin showed some small net loss of anthocyanin color and an increase in net browning, indicative of interaction, whereas effects were less marked with mixtures with procyanidins B2 and C1. The net loss of Mv 3-glucoside in the mixture was confirmed by acidifying to ph <1, when the color was only 88 % of that found on acidification of the anthocyanin alone. Solutions were examined at intervals from the beginning of storage by chromatography (paper and thin-layer). After seven weeks, little interaction was apparent between Mv 3-glucoside or Mv 3,5- diglucoside and procyanidin B2 since the anthocyanins appeared largely unchanged. Remarkably, B2 had been partly transformed or disproportionated into epicatechin and trimer C1, but that occurred also in the control solution containing B2 alone INTERACTIONS BETWEEN ANTHOCYANINS, PHENOLIC COMPOUNDS, AND ACETALDEHYDE--99 (Fig. 1). When interaction is slow as under those conditions, it appears that procyanidins such as B2 can undergo further transformations much faster than their direct interaction with anthocyanins. After seven months of storage, solutions containing Mv 3-glucoside and epicatechin or B2 still exhibited well-defined spots of the original components but, particularly noticeable by TLC, the phenolic degradation pattern was different from that of the phenolic alone, suggesting that some interaction had occurred. More significantly, visibly yellow spots appeared, some mobile in s-baw but not in HOAc, and one mobile in both solvents (Fig. 1). As described later herein, these new spots respectively correspond to the xanthylium compounds X1 and X3. During the seven months of storage the control phenolic solutions browned in direct proportion to their molecular weights, i.e., identically on a weight basis. Chromatography showed that C1 as well as B2 had undergone further transformations, both in the control phenolic solutions and also in admixture with the anthocyanin. The dimer B2 showed the same disproportionation pattern as earlier, and trimer C1 had formed appreciable amounts of epicatechin, B2, and more polymeric procyanidins, the last being characterized by their immobility in s-baw and streaking in HOAc. The changes in B2 under similar conditions were examined in more detail. Thus, B2 dissolved in alcoholic tartrate buffer (ph 3.5) and left in test tubes at laboratory temperature for some weeks was partly transformed into epicatechin and trimer C1, with small amounts of dimer B5 (19) and more complex procyanidins. The disproportionation occurred to only a small extent in darkness but was greatly accelerated by light. Exposure to light also caused increased production of a yellow-brown chromophore (hm~x 450-460 rim), further accelerated by adding iron salts. The transformations of the procyanidins could not be reproduced in acetate buffer (ph 3.5; without added iron) on light exposure, even in the presence of traces of acetaldehyde (thought to be a possible reactant), and the reason for the differences obrained according to buffer anion are as yet unexplained. Solutions containing Mv 3,5-diglucoside and various phenolics were examined after one year. They exhibited an intense browning (hm~x 440-450 rim) ~vhich appeared much more pronounced than that occurring with Mv 3-glucoside because of the comparative lack of red color as a result of the lower pk value of the diglucoside (28). Browning of corresponding solutions containing Mv 3-p-coumarylglucoside-5-glucoside was intermediate between mono- and diglucosides. Chromatography of solutions containing the diglucoside or acylated diglucoside and (+) -catechin or epigallocatechin revealed discrete yellow spots which turned orange on exposure to ammonia vapor (Fig. 1). The components responsible were worked up in HOAc, giving three bands of increasing Rf, viz., X1 (immobile), X2,

100--INTERACTIONS BETWEEN ANTHOCYANINS, PHENOLIC COMPOUNDS, AND ACETALDEHYDE 20-1.0-! I! I I ' 400 500 600 WAVELENGTH (nm) Fig. 2. Colors of model systems (tartrate buffer; ph 3.5; 2 mm cell-length). Curve 1. Mv 3-glucoside alone at start; 2. Mv 3-glucoside alone after 198 days; 3. Mv-glucoside + (+)- catechin after 198 days; 4. Mv 3-glucoside + acetaldehyde after 198 days; 5. Mv 3-glucoside + acetaldehyde + (+)-catechin after 19 days. and X3. When purified further, their spectral characteristics were in agreement with those expected for xanthylium salts (11,17). as described later. Their Rf values indicated X1 and X2 to be aglycones, with X3 glycosylated. Bands X1 and X3 correspond to those detected previously on the chromatograms of the Mv 3-glucoside solutions. Partial hydrolysis of the acylated diglucoside to Mv 3,5-diglucoside and p-coumaric acid was also observed (Fig. 1). Reaction of anthocyanins with iron: There was little net effect of ferric chloride on Mv 3-glucoside in darkness but the color was augmented and became violet on exposure to light. The latter effect can be attributed to the production of reactive carbonyl compounds by the well-known photochemical decomposition of tartaric acid in the presence of oxygen and iron salts (20), and was similar to the direct effect of acetaldehyde described later herein. In contrast, the effect on Mv 3,5-diglucoside in light was negligible. Reaction of anthocyanins with acetaldehyde: Acetaldehyde gradually produced a small color augmentation with Mv 3-glucoside, accompanied by a change in hue to violet (Fig. 2). One such solution soon formed a precipitate, but another remained clear and after seven months still gave its original spot on chromatography although with other degraded material. After this time the color was augmented by 60%, but the color found on acidification was similar to that without acetaldehyde. After 8-9 months the solution remained violet but contained 5 largely chromatographically immobile and degraded material, with little detectable monomeric anthocyanin. In contrast, acetaldehyde was largely without effect on Mv 3,5-diglucoside and Mv 3-p-coumarylglucoside-5-glucoside. Garoglio (7) previously found similar differences between anthocyanidin 3-mono and 3,5-diglucosides. Reaction of anthocyanins with phenolics and acetaldehyde: a) My 3-glucoside: The color o solutions increased rapidly and continuously and became violet in the presence of acetaldehyde and the following phenolics, epicatechin, ( + )-catechin, procyanidins B2 and C1, and crude samples containing epigallocatechin, gallocatechins, and gallocatechin gallates (Fig. 2). Quercetin rhamnoside had only a small effect. Interactions with epicatechin and the procyanidins were followed in detail. With epicatechin, a new anthocyanin, chromatographically distinguishable from Mv 3-glucoside, was formed within 2 days (spot ME, Fig. 1). After seven days, when the color was almost doubled, it became the major anthocyanin under some experimental conditions; thereafter precipitation began. In the presence of B2, the color increase was twice as rapid and a new anthocyanin was again detected within 2-4 days before precipitation started. With C1, the color increase approached the rate of B2 but no new anthocyanin was observed before precipitation occurred. Examination of the phenolics alone plus acetaldehyde showed that epicatechin was relatively 1.0. 0.8-0,3-0.4-0.2- '0 0 0 0 0 0 0 C10 0 0-0 ME~ T O MBO i I i? v i :," 1 2 3 4 5 6 7 8 Fig. 3. Thin-layer chromatography of components. 1. (-)-epicatechin (E) standard; 2. procyanidin dimer B2 standard; 3. procyanidin trimer C1 standard; 4. Mv 3-glucoside (M) 5. Mv 3- glucoside + epicatechin + acetaldehyde after 7 days (ME is the new compound); 6. Mv 3-glucoside + B2 + acetaldehyde after 4 days (MB is the new compound); 7. B2 after 8 weeks; 8. C1 after 7 months (T is probably a tetrameric procyanidin).

INTERACTIONS BETWEEN ANTHOCYANINS, PHENOLIC COMPOUNDS, AND ACETALDEHYDE--101 stable (up to 22 days), that B2 was degraded completely within 22 days, possibly via an intermediate transformation of the type already described, and that C1 was degraded even more rapidly. The formation of the new anthocyanins was verified by TLC using the silica gel system, which broadly gives decreasing R~ values for phenolics of increasing complexity, e.g., epicatechin > B2 > C1. It was noticeable that the R~ value of the new anthocyanin produced by interaction with B2 was lower than that formed from epicatechin (spots MB and ME, Fig. 3). Both separated well from the unchanged Mv 3-glucoside of highest R~. Further, on paper chromatograms the new anthocyanins gave positive phenolic reactions with diazotized p-nitroaniline, a standard reagent for detection of phenolics. The findings are thus consistent with the supposition that Mv 3-glucoside can combine with epicatechin and with procyanidin B2, probably via CH3CH linkages, to give previously unidentified compounds which behave like anthocyanins and phenolics. The formation of similar compounds in ports has been postulated earlier (22) although discrete entities were not demonstrated. The failure of quercetin rhamnoside to undergo similar reactions can be accounted for by the deactivating effect of its carbonyl group on the reactive 8 and 6 positions of its A ring. Further tests carried out with (+)-catechin showed that the reaction occurred readily between ph 2.75 and 4.00, with an optimum at ph 3.75-4.00. Here the absorbance doubled before precipitation occurred. b) Mv 3-p-coumarylglucoside-5-glucoside: The color intensity increased gradually and became violet, similar to the behavior of Mv 3-glucoside but more pronounced, in the presence of (+)-catechin and crude epigallocatechin. Unlike those of the 3- glucoside, however, solutions remained stable without precipitation for up to 40 days. A new anthocyanin was detected in the solutions containing + )-catechin (spot A-cat, Fig. 1). c) Mv 3,5-diglucoside: Mv 3,5-diglucoside reacted similarly with the phenolic compounds already mentioned for the 3-glucoside but with the following differences. It gave greater color, which reached a maximum and was then stable without precipitation, also the largest difference in Rf (in HOAc) between the newly formed anthocyanin (e.g., spot D-cat, Fig. 1) and unchanged diglucoside, thus offering the best possibility of separation. The phenolic used for subsequent work was principally (+)-catechin. The color augmentation was unaffected by the nature of the buffer anion (acetate, succinate, malate, tartrate, or citrate) and it occurred readily from 2.75-4.00. The optimum ph for reaction appeared about 3.5, as estimated by the greatest percentage increase in absorbance. The comparative stability of the newly formed anthocyanin (s) was emphasized by measurements made over an extended period. At ph 3.5, color reached 2.O- 1.O- I I I I I 400 500 60 O WAVELENGTH (nm) Fig. 4. Colors of model systems (tartrate buffer; ph 3.5; 10 mm cell-length). Curve 1. Mv 3,5-diglucoside alone at start; 2. Mv 3,5-diglucoside alone after 198 days; 3. Mv 3,5-diglucoside + (+)-catechin after 163 days; 4. Mv 3,5-digiucoside + acetaldehyde after 198 days; 5. Mv 3,5-diglucoside + (+)-catechin + acetaldehyde after 19 days (the color of this solution increased further to 5 times its original color after 57 days). a maximum after 57 days when the color was 5 times its original (Fig. 4). This contrasts with 3 times color augmentation of the acylated diglucoside and only doubling of the color of the monoglucoside before precipitation occurred. At ph values lower than 3.5 the color of the diglucoside solution then decreased with further storage whereas at higher ph values it increased slowly, partly because of browning. The new anthocyanin (spot D-cat, Fig. 1) was still detectable after five months at ph 3 although accompanied by considerable degraded material and unchanged diglucoside, but was less evident at ph 4. Examination of corresponding mixtures in acetate rather than tartrate buffer after seven months revealed the new anthocyanin and also yellow components, the major one being mobile in both solvents. A small amount of the latter was isolated and was identified as a glucosylated derivative of a xanthylium salt by its spectral and chromatographic similarities to the glucose derivative of a 1,3,6,8-tetrahydroxyanthylium salt previously found in stored grape juice (11, Table 1). Heating in acid (N HCI) resulted in decreased Rf values as expected for its hydrolysis to the aglycone. The glucosylated component corresponded to X3 formed from Mv 3-glucoside. The simultaneous occurrence of both xanthylium salts and the new anthocyanins induced by acetaldehyde indicates that both reactions can occur separately and by independent mechanisms. Isolation and properties of the new anthocyanins: The new anthoeyanins were isolated by preparative paper chromatography from a concentrated solution of Mv 3,5-diglucoside, acetaldehyde, and (+)-carechin. Three fractions were obtained of increasing

102-INTERACTIONS BETWEEN ANTHOCYANINS, PHENOLIC COMPOUNDS, AND ACETALDEHYDE Rf values in s-baw; they were more violet than Mv 3,5-diglucoside. The most mobile fraction III was further separated in HOAc into the new anthocyanin III(1) and unchanged diglucoside III(2) (Table 2). Equal weights (ca. 2.5 mg) of every fraction were dissolved in tartrate buffer (ph 3.5; Table 1. Identification of glycosylated xanthylium salt (X3). Solvent X3 Glycosylated xanthylium salt (11) 20-'. Spectral peaks 1% HCI 271, 304(mi), 439 (Xmax nm) BuOH-HCI 272, 308(mi), 456 NaOH 280, 465(sh), 489 269, 302(mi), 439 272, 305, 454 273, 465(sh), 486 10- R~ values A 0.53 0.68 B 0.80 0.76 mi- minor; sh - shoulder C 0.63 0.76 = = J i = ---"- -T 400 5O0 6O0 20 - WAVELENGTH (nm) Table 2. Components formed by reaction of Mv 3,5-diglucoside with (+)-catechin and acetaldehyde. ph 3.5 ph < 1 % formed Xmax 10mm a ~max 10mm Fraction by weight (nm) EX max (nm) EX max Degree of ionisation at ph 3.5(26) 10- I 25 too 0.20 weak II 36 553 0.64 III (1) 10 555 2.33 III (2) (Mv 3,5 -diglucoside) 0.32 4.74 62 537 1.44 45 536 29 535 0.61 524 11.38 5 a Concentrations approx. 0.5 mg/ml (see text). 49 /! i 400 5~)O 6~)O ' WAVELENGTH (nm) Fig. 5. Variation of anthocyanin color with ph. (a) Mv 3,5- diglucoside - acetaldehyde - (+)-catechin compound; (b) Mv 3,5-diglucoside. 1. 0.5N HCI; 2. ph 2.74; 3. ph 4.51; 4. ph 5.41; 5. ph 6.85 (buffers 2-4 were citrate-phosphate; buffer 5 was phosphate; all solutions were allowed to equilibrate for 24 hours). Table 3. Spectral and Rf data of new anthocyanin from Mv 3,5-diglucoside, acetaldehyde and (+)-catechin. Spectra a Rf values Euv.max. E440 Paper TLC max Evis.max. Evis.max. HOAc HOAc-HC1 HOAc b HOAc (nm) (%) (%) New anthocyanin 280, 548 Mv 3,5-277, 540 diglucoside 72 19 0.27 0.58 0.07 0.03 49 11 0.52 0.51 0.37 0.20 a Methanol containing 0.01% HCI. b Cellulose. c Cellulose 80:PVP 20 (33).

5 ml), and the absorbances of the solutions were measured before and after acidification to ph < 1. Acidification only doubled the absorbances of the new fractions compared with a twenty-times increase in that of Mv 3,5-diglucoside (Table 2). Fraction III (1) was very much more colored at ph 3.5 than Mv 3,5-diglucoside, and was examined in more detail. Its spectral and Rf values are contrasted with those of Mv 3,5-diglucoside in Table 3 and Fig. I (spot D-cat). The increased absorbance at 280 nm is consistent with the presence of an equimolar amount of catechin which itself gives a peak at 280 nm. Its Rf values in s-baw and BAW were respectively slightly lower and slightly higher than those of Mv 3,5-diglucoside. Color variations with ph are contrasted in Fig. 5 (a) and (b). The new anthocyanin exhibits appreciable color at ph values (2.7-6.9) where the diglucoside is virtually colorless, probably through stabilization of its anhydrobase. Although we have not yet categorically shown the presence of catechin in the new anthocyanin it seems likely that it is linked via acetaldehyde to the Mv 3,5-diglucoside. By analogy with the behavior of 7-hydroxy, 7-hydroxy-6-methyl, and 7-hydroxy-8- methyl flavylium chlorides, the linkage is probably to the anthocyanin 8 position. Thus, introduction of a methyl group into position 6 of 7-hydroxy flavylium chloride does not alter the peak position (}'m~x) and has little effect on anhydrobase formation. Compared with the 7-(OH)-flavylium salt, however, 7- (OH) -8- (CH3) -flavylium chloride exhibits a bathochromic shift of 8 nm, is much more colored at ph 2-4 and forms its anhydrobase at a much lower ph. DISCUSSION Studies in model wine systems, consisting of alcoholic tartrate buffer containing pure anthocyanins. based on malvidin, phenolic compounds based on the catechins, and acetaldehyde, alone and in admixtures, have shown that at least six reactions can occur which have implications in wine-making. They are summarized and discussed in the following order" 1) Reactions between anthocyanins and phenolics. 2) Transformations of phenolics alone. 3) Degradation of anthocyanins alone. 4) Reactions between anthocyanins and acetaldehyde. 5) Interactions of anthocyanins, acetaldehyde, and phenolics. 6) Reactions between phenolics and acetaldehyde. Reaction 1 involves a very slow condensation between anthocyanin and phenolic; the eventual formation of xanthylium salts (as depicted in Fig. 6) indicates that the linkage must be between position 4 of the anthocyanin and position 8 (or 6) of the phenolic. Xanthylium salts were formed from all three anthocyanins examined, and their formation was not noticeably slower from the diglucosides INTERACTIONS BETWEEN ANTHOCYANINS, PHENOLIC COMPOUNDS, AND ACETALDEHYDE--103 (11) OGI HO 6 + OH O OCH 3 OH OH H O ~ ",.-. Fig. 6. Formation of xanthylium salts (111) from Mv 3,5-diglucoside (I) and (+)-catechin (11). (The precise formulae of III have yet to be determined). (acylated and nonacylated) than from the monoglucoside. It thus appeared that hydrolysis of the glucose in the anthocyanin position 5 (Fig. 6) was not rate-determining and that it could occur more readily than hydrolysis of the glucose in position 3, which must have remained in the glycosylated xanthylium salt (X3). A precedent already exists for ready loss of a substituent in position 5, e.g., a methyl group during the transformation into xanthylium salts of the condensation product of 5,7, 31,41-tetramethoxy-flavan-3,4-diol with phloroglucinol (17). There was some evidence that Reaction 1 might occur more readily with the monomeric catechins rather than the procyanidins but the conclusions were obscured by Reaction 2, in which the latter disproportionated into epicatechin and lower and higher condensed procyanidins. It is known that the C-C link in procyanidins is labile and readily hydrolyzed by acids and other reagents (27), but it is remarkable that both hydrolysis and formation of higher polymers should occur under such mild conditions. The findings have implications concerning the occurrence and biosynthesis of the procyanidins, which are beyond the scope of the present article. Superimposed upon these subtle transformations was the oxidative browning of the phenolics which was catalyzed by iron and light exposure. The spontaneous degradation of the anthocyanins alone (Reaction 3) was the least prominent reaction occurring without acetaldehyde. The monoglucoside appeared marginally less stable than the diglucosides since it lost most red color (Fig. 2, curve 2). All three anthocyanins showed some browning during this time. Addition of acetaldehyde to solutions containing anthocyanin alone gave a reaction only when the latter was Mv 3-glucoside (Reaction 4). Acetaldehyde probably reacts, as an electrophile, at the nucleophilic 8-or 6-sites of the anthocyanin A ring, with the formation of complexes consisting of mono- HO (111)

104--INTERACTIONS BETWEEN ANTHOCYANINS, PHENOLIC COMPOUNDS, AND ACETALDEHYDE H H I I CH 3- C = 0 + H* > CH3- C -OH (v) (IV) ~ +(11) CHiC-OH II.I cat + H + HO.~,,,-O~ H20 + CH 3- C-H ~ o. (VII) [ OH +(I) (v0 cat I/~OH OCH 3 cat K~OH OCH3 HO CH~'C-H 'J" " 1" ~ 11 D //'OCH3 [ O CH~C-H ~, ~ O ~ 'l _1 OCH 3 H +4- "Tf/A I~ +,:T v. _ H + OGI (viii) OGI OH (ix) + H + Fig. 7. Reaction of acetaldehyde (IV) with (+)-catechin (11) and Mv 3,5-diglucoside (I). glucoside units linked by CH3CH bridges. Resultant stabilization of the anhydrobase forms of these complexes accounts for their violet color. The diglucoside and acylated diglucoside were unaffected because of the reduced nucleophilic activity at the reactive positions as a result of electron withdrawal by the glucose in position 5. The latter effect is well-known also to result in an increase in the net positive charge (electrophilic activity) at positions 4 and 2 of the heterocyclic ring of diglucosides compared with monoglucosides, thus accounting for the greater acidity (lower pk values) of the former. The same considerations lead one to expect diglucosides to condense more readily than monoglucosides with phenolics, implying that the former may be more significant than the latter during aging of hybridwines, although no definite evidence was obrained on that point in these experiments. In the presence of both catechin-type phenolics and anthocyanins, rapid reaction (Reaction 5) ensued on addition of acetaldehyde, with the formation of new violet anthocyanins of great color intensity. Malvidin 3-glucoside was most reactive under these conditions and soon formed complexes large enough to precipitate from solution. In contrast, the 3,5-diglucosides were less reactive (so permitting the more ready isolation of intermediates), again because of the electron-withdrawing properties of the 5-glucose. The reaction can be envisaged according to Fig. 7. Acetaldehyde (IV) reacts as its carbonium ion (V) with catechin (II), probably at position 8 (9) rather than 6, giving compound (VI). The latter forms a further carbonium ion (VII), with loss of water, which reacts with the anthocyanin (I), probably at position 8. The resultant compound (VIII) is then readily stabilized by deprotonation to the violet quinonoid or anhydrobase (IX). Linkage is considered to be more likely to position 8 of the anthocyanin rather than position 6, by analogy with the behavior of the simple flavylium salts and also because of the higher net negative ground state charge of position 8 reported for cyanidin (3). The carbonium ion of the substituted catechin (VII) is likely to be stabilized, compared with (V), by the adjacent phenyl groups of catechin. As such, it will react more readily with anthocyanin than (V), thus accounting for the much faster reaction of acetaldehyde with both phenolic and anthocyanin than with anthocyanin alone. The rate of Reaction 5 appeared to increase with complexity of the phenolic, i.e., epicatechin < B2 = C1. A similar order of reactivity was observed during the reaction of phenolic alone with acetaldehyde (Reaction 6), but, here again, the results were complicated by simultaneous disproportionation of the procyanidins. No discrete intermediates were observed, as in the presence of anthocyanins, but rather streaking or reduced mobility in both chromatographic solvents. At sufficiently large phenolic concentration precipitation soon occurred. It is likely that the complexes formed contained phenolic units linked by acetaldehyde bridges, much as described previously. Since reaction may occur at position 6 as well as 8, in mixed systems containing both a phenolic (Ph) and an anthocyanin (A) there may be present not only regular or sequential polymers of varying molecular size, e.g., Ph-CH (CH3)-A; Ph-CH (CH3)- A-CH (CH3)-Ph, etc., but also complexes containing irregular substitution of components such as Ph- CH (CH3)-Ph-CH (CH~) etc. Such structures may have been present in the less mobile bands found during isolation of the new anthocyanin (Table 2). When the phenolic and anthocyanin can also vary in nature, as in red wines which contain many different components, it is evident that a very complex situation could exist. Observations on the effect of acetaldehyde addition to wines have been made for a number of years. In the most recent work (4), color was enhanced not only by acetaldehyde but also by invert sugar. In view of this report we have to state that adding pure aldehyde-free glucose or fructose to our model systems produced no effect comparable to that of acetaldehyde. Acetaldehyde is formed in wines by several mechanisms, principally by microbial action during fermentation and more slowly from ethanol by coupled oxidation of certain phenolics (32). While the amount of acetaldehyde found in wines is much less than the large excess which was deliberately added to our model systems, it is likely that similar reactions would occur, albeit more slowly. Indeed it is reported (1) that the acetaldehyde content is usually lower in red wines than in white wines because of the occurrence of phenolics and anthocyanins in the former. The interactions could be of the type described here, which well account for the purple tints observed in young wines. The extent of the reactions, i.e., whether red

wine color is augmented or eventually diminished (by prolonged condensation leading to precipitation) must depend upon the relative concentrations of the components involved as well as their nature. In white wines all the free acetaldehyde is available for reaction with phenolics, with consequent deleterious effects on color, flavor, and quality. In red wines, acetaldehyde initially augments the color, and the implications on quality are being currently investigated following recent work relating color and quality (26). The more stable augmentation of diglucoside color may partly account for the strong color of hybrid wines, which on consideration of pk values might be expected to be less colored. Sulfur dioxide is the most significant agent which determines the amount of free acetaldehyde in wines. It perhaps needs emphasizing that the appearance of free acetaldehyde is a critical stage in winemaking, heralding a series of new reactions. If these are to be avoided it is necessary to ensure a small but measurable (5) content of free sulfur dioxide in the wine. The direct condensation of anthocyanins and phenolics (without acetaldehyde intervention) appeared much slower in model systems, even in air, than the reported rate of formation of polymeric pigments in an actual wine (24). The operation of additional factors during red wine aging, which have not yet been considered in model systems, is thus indicated; these could well include oxidizing enzymes in the pulp. The finding of the yellow xanthylium chromophores in model systems lends further support to their formation also in the polymeric wine pigments as proposed previously (16,17). LITERATURE CITED 1. Amerine, M. A., and M. A. Joslyn. Table Wines. 2nd ed. pp 449-50. University of California Press, Berkeley and Los Angeles. (1970). 2. Asen, S., R. N. Stewart, and K. H. Norris. Co-pigmentation of anthocyanins in plant tissues and its effect on colour. Phytochemistry 11:1139-44 (1972). 3. Bendz, G., O. Martensson, and E. Nilsson. Studies of flavylium compounds. I. Some flavylium compounds and their properties. Arkiv fl~r Kemi 27:65-77 (1967). 4. Berg, H. W., and M. A. Akiyoshi. On the nature of reactions responsible for color behavior in red wine: a hypothesis. Am. J. Enol. Vitic. 26:134-43 (1975). 5. Burroughs, L. F. Determining free sulfur dioxide in red wine. Am. J. Enol. Vitic. 26:25-9 (1975). 6. Friedrich, H.,and J. Schonert. Separation of flavans from Vaccinium m!jrtieus. Archiv der Pharmazie 306:611-18 (1973). 7. Garoglio, P. G. Proposta di modifiche nella manualita del metods rapids Dorier e Verelle per la determinazione del diglucoside malvosidico nei vini. Riv. Viticolt. Enol. 21:11-15 (1968). 8. Haslam, E. Natural Procyanidins. Chapter 10 in "The Flavonoids". (J. B. Harborne, T. J. Mabry, and H. Mabry eds.) pp 505-59 (1975). 9. Hillis, W. E., and G. Urbach. The reaction of (+)-catechin with formaldehyde. J. Appl. Chem. 9:474-82 (1959). 10. Hrazdina, G. Reactions of anthocyanins in food products. Lebensmitt.-Wiss.u.Technoi. 7:193-8 (1974). INTERACTIONS BETWEEN ANTHOCYANINS, PHENOLIC COMPOUNDS, AND ACETALDEHYDE--105 11. Hrazdina, G., and A. J. Borzell. Xanthylium derivatives in grape extracts. Phytochemistry 10:2211-13 (1971). 12. Hrazdina, G., A. J. Borzell, and W. B. Robinson. Stability of the anthocyanin 3,5-diglucosides. Am. J. Enol. Vitic. 21:201-4 (1970). 13. Jurd, L. Anthocyanidins and related compounds. VI. Flavylium salt-phloroglucinol condensation products. Tetrahedron 21:1471-83 (1965). 14. Jurd, L. Anthocyanidins and related compounds. XI. Catechin-flavylium salt condensation reactions. Tetrahedron 23:1057-64 (1967). 15. Jurd, L. Anthocyanidins and related compounds. XVI. The dimerization of flavylium salts in aqueous solutions. Tetrahedron 28:493-504 (1972). 16. Jurd, L. Review of polyphenol condensation reactions and their possible occurrence in the aging of wines. Am. J. Enol. Vitic. 20:191-5 (1969). 17. Jurd, L., and T. C. Somers. The formation of xanthylium salts from proanthocyanidins. Phytochemistry 9:419-27 (1970). 18. Lea, A. G. H. The isolation and quantitative determination of phenolic compounds in cider. M.Sc. thesis, University of Bristol, p. 30 (1974). 19. Lea, A. G. H., and C. F. Timberlake. The phenolics of ciders. I. Procyanidins. J. Sci. Fd. Agr. 25:1537-45 (1974). 20. Pollard, A., and C. F. Timberlake. Fruit juices. Chapter 17 in "The biochemistry of fruits and their products". (A. C. Hulme, ed.). Vol. 2. Academic Press, London and New York. pp 573-621 (1971). 21. Rib~reau-Gayon, P. Interpretation chimique de la couleur des vin rouges. Vitis 12:119-42 (1973). 22. Singleton, V. L., H. W. Berg, and J. F. Guymon. Anthocyanin color level in port type wines as affected by the use of wine spirits containing aldehydes. Am. J. Enol. Vitic. 15:75-81 (1964). 23. Somers, T. C. Pigment profiles of grapes and wines. Vitis 7:303-20 (1968). 24. Somers, T. C. The polymeric nature of wine pigments. Phytochemistry 10:2175-86 (1971). 25. Somers, T. C. Wine tannins--isolation of condensed flavonoid pigments by gel-filtration. Nature 209:368-70 (1966). 26. Somers, T. C., and M. E. Evans. Wine quality: correlations with colour density and anthocyanin equilibria in a group of young red wines. J. Sci. Fd. Agr. 25:1369-79 (1974). 27. Thompson, R. S., D. Jacques, E. Haslam, and R. J. N. Tanner. Plant proanthocyanidins. Introduction, the isolation, structure and distribution in nature of plant procyanidins. J. Chem. Soc. Perkin Transactions I, 11:1387-99 (1972). 28. Timberlake, C. F., and P. Bridle. Flavylium salts, anthocyanidins and anthocyanins. I. Structural transformations in acid solutions. J. Sci. Fd. Agr. 18:473-8 (1967). 29. Timberlake, C. F., and P. Bridle. Flavylium salts resistant to sulphur dioxide. Chemistry and Industry 1489 (1968). 30. Timberlake, C. F., and P. Bridle. The effect of processing and other factors on the colour characteristics of some red wines. Vitis 15:37-49 (1976). 31. Weinges, K., and M. V. Piretti. Zur Kenntnis der Proanthocyanidine, XY~lsolierung des C30H2~O~2--procyanidin B1 aus Weintrauben. Liebigs Ann. Chem. 748:218-20 (1971). 32. Wildenradt, H. L., and V. L. Singleton. The production of aldehydes as a result of oxidation of polyphenolic compounds and its relation to wine aging. Am. J. Enol. Vitic. 26" 25-9 (1974). 33. Wrolstad, R. E. Thin layer chromatography of anthocyanins on mixed layers of PVP and cellulose. J. Chromatog. 37" 542-4 (1968).