Racemization of Aspartic Acid and Phenylalanine in the Sweetener

Proc. Natl. Acad. Sci. USA Vol. 81, pp. 5263-5266, August 1984 Chemistry

Racemization of aspartic acid and phenylalanine in the sweetener aspartame at 1000C (aspartame racemization kinetics/D amino acids/diketopiperazine racemization) MARCUS F. BOEHM AND JEFFREY L. BADA* Amino Acid Dating Laboratory (A-012B), Scripps Institution of Oceanography, University of California at San Diego, La Jolla, CA 92093 Communicated by Stanley L. Miller, April 23, 1984

ABSTRACT The racemization half-lives (i.e., the time re- basic pH values. Since the presence of D amino acids in the quired to reach a D/L = 0.33) at pH 6.8 for aspartic acid and human diet may have a variety of consequences, many of phenylalanine in the sweetener aspartame (L-aspartyl-L-phen- which are only poorly understood (11-13), it would be im- ylalanine methyl ester) were determined to be 13 and 23 hours, portant to establish the rate of racemization of aspartic acid respectively, at 100'C. Racemization at this pH does not occur and phenylalanine in aspartame and its diketopiperazine de- in aspartame but rather in its diketopiperazine decomposition composition product. As far as we could ascertain, no previ- product. Our results indicate that the use of aspartame to ous aspartame racemization studies had been reported. sweeten neutral pH foods and beverages that are then heated Thus, we have investigated the racemization kinetics of the at elevated temperature could generate D-aspartic acid and D- aspartic acid and phenylalanine in solutions of aspartame phenylalanine. The nutritive consequences of these D-amino heated at 100°C and pH 4 and 7. acids in the human diet are not well established, and thus aspartame should probably not be used as a sweetener when the exposure of neutral pH foods and beverages to elevated MATERIALS AND METHODS temperatures is required. At pH 4, a typical pH of most foods The pure L-aspartyl-L-phenylalanine methyl ester was ob- and beverages that might be sweetened with aspartame, the tained from Sigma. The low-caloric sweetener Equal half-lives are 47 hours for aspartic acid and 1200 hours for (Searle), which contains aspartame, cellulose, and corn phenylalanine at 100'C. Racemization at pH 4 takes place in sweeteners, was obtained from a local supermarket. All oth- aspartame itself. Although the racemization rates at pH 4 are er chemicals used were the purest grade commercially avail- slow and no appreciable racemization of aspartic acid and ble. phenylalanine should occur during the normal use of aspar- Succinate (pH 4) and phosphate (pH 6.8) buffer solutions tame, some food and beverage components could conceivably containing either 0.01 M aspartame or 0.01 M Equal were act as catalysts. Additional studies are required to evaluate prepared. The total buffer concentrations were 0.1 M. The whether the use of aspartame as a sugar substitute might not in final ionic strength of the solutions was adjusted to 0.5 M turn result in an increased human consumption of D-aspartic with NaCl. Aliquots of the solutions were sealed in glass acid and D-phenylalanine. tubes and placed in a constant-temperature 100°C heating block for various lengths of time. After heating, the samples A reduction in human sugar consumption is advocated for were dried, hydrolyzed in 6 M HCl for 6 hr at 100°C, and numerous reasons, with the prevention of dental caries and then desalted on Dowex 50W-X8 cation-exchange resin. As- obesity probably being the most compelling. However, ef- partic acid and phenylalanine were eluted from the resin with forts to find a sugar substitute have not been generally suc- 1-2 M NH40H. cessful. Recently, a new sweetener, aspartame, the methyl To compare the extent of racemization in aspartame with ester of the dipeptide L-aspartyl-L-phenylalanine, has been its diketopiperazine decomposition product, the solution approved by the Food and Drug Administration as a sugar heated for 24 hr was chromatographed on Dowex 50W-X8 substitute in certain foods and beverages. Aspartame has a cation-exchange resin (7). The diketopiperazine fraction was sweetness factor nearly 200 times that of sucrose (1, 2). obtained by collecting the water eluant from the cation col- Moreover, since aspartame is composed of amino acids it is umn. Aspartame was then eluted from the column with 1.5 classified as a "natural" sweetener in contrast to saccharin, M NH40H. The aspartame and diketopiperazine fractions which is considered artificial (1). Whether aspartame is a were hydrolyzed in 6 M HCI for 6 hr at 100°C and then de- safe alternative to sugar in the human diet, however, remains salted as described above. controversial (3, 4). Several potential problems exist. For The N-trifluoroacetyl-L-prolyl chloride reagent was used example, some individuals (phenylketonurics) are highly for synthesizing amino acid derivatives for gas chromato- susceptible to excess phenylalanine in their diet and thus are graphic based enantiomeric analyses. This reagent was pre- cautioned on the package of the commercially available pared according to the procedures described elsewhere (14). product about using aspartame. A portion of the dried extract from the hydrolyzed, desalted In aqueous solution, aspartame decomposes (1, 5, 6) via a samples was dissolved in 1 ml of HCl/methanol and heated series of reactions that include ester and peptide-bond hy- at 100°C in order to synthesize the amino acid methyl esters. drolysis and cyclization to the diketopiperazine, 3-carboxy- The methyl esters were then dissolved in 1 ml of dichloro- methyl-6-benzyl-2,5-piperazinedione. We were particularly methane and allowed to react with 1 ml of N-trifluoroacetyl- interested in the diketopiperazine pathway since results ob- L-prolyl chloride along with 5 drops of triethylamine in order tained in this laboratory (7) and in earlier work (8-10) had to synthesize the N-trifluoroacetyl-L-prolyl amino acid demonstrated that amino acid residues in diketopiperazines methyl esters. This derivatization was carried out at room are highly susceptible to racemization at both neutral and temperature for 1 hr after which 1 ml of 6 M HCl was added to terminate the reaction. The solution was then stirred and The publication costs of this article were defrayed in part by page charge centrifuged to separate the organic and aqueous layers. The payment. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. §1734 solely to indicate this fact. *To whom reprint requests should be addressed. 5263 5264 Chemistry: Boehm and Bada Proc. Natl. Acad Sci. USA 81 (1984) organic layer containing the diastereomeric dipeptide methyl rapidly than phenylalanine. At pH 6.8, however, aspartic esters was removed and dried over magnesium sulfate. acid and phenylalanine have racemization rates that are Gas chromatography was carried out on a fused silica cap- nearly equivalent. Also, the racemization rates in aspartame illary column (30 m x 0.25 mm) coated with OV 101 using a at pH 6.8 exceed those at pH 4. In experiments using the Hewlett-Packard model 5710A gas chromatograph equipped commercial product Equal the same trends are apparent. with a splitter. The flow rate was about 3.0 cm3-min-1. The The rates at pH 6.8 are slower, however, than we deter- gas chromatography conditions were as follows: injection mined in the experiments with pure aspartame, whereas port, 200'C; detection port, 250'C; initial column tempera- those at pH 4 are roughly similar to the pure aspartame val- ture, 120'C. The oven temperature was increased at a rate of ues. 40C per min and was finally held at 220'C for 8-16 min. Un- The racemization half-lives for aspartic acid and phenylal- der these conditions the retention time of the aspartic acid anine in aspartame at 100°C are summarized in Table 1; also and phenylalanine diastereomeric derivatives is about 26-28 included are the half-lives in other systems. At pH 4, the min and 36-38 min, respectively. The enantiomers of both racemization half-lives we have determined for aspartic acid amino acids were baseline separated. The D/L aspartic acid and phenylalanine in aspartame are nearly the same as the (D/L Asp) and D/L phenylalanine (D/L Phe) ratios were cal- values for the free amino acids at this pH. At pH 6.8, howev- culated from peak heights. er, aspartic acid and phenylalanine in aspartame are race- The aspartic acid D/L ratios were also determined by using mized much more rapidly than the free amino acids. In the diastereomeric dipeptide method (15). The dipeptides L- chicken muscle heated at 121°C, aspartic acid undergoes ra- leucyl-L-aspartic acid and L-leucyl-D-aspartic acid were syn- cemization faster than in aspartame at 100°C. On the other thesized as described elsewhere (15, 16) and then separated hand, phenylalanine in aspartame has a racemization rate at on a Beckman model 119C amino acid analyzer interfaced 100°C and pH 6.8 that is much faster than that in chicken with a DEC PDP11 computer, which was used for data ac- muscle at 121°C. quisition and analysis. The D/L Asp ratios obtained by this method and the gas chromatographic method described DISCUSSION above agreed within ±5%. Since our racemization measurements are carried out by us- RESULTS ing the total aspartic acid and phenylalanine in the heated Amino acid racemization obeys (17, 18) the rate equation: solutions, the rates that we have determined are actually the composite values for racemization not just in the dipeptide 1 + D/L but also in its formed In D/ - In1+ =2*krt, aspartame decomposition products 1 D/L - D/L during the heating experiments. ) )'=° Aspartame decomposition at neutral pH has been found to in which ki is the first-order rate constant for interconversion be much more rapid than at acidic pH values, with the major of the enantiomers and D/L is the enantiomeric ratio at time product being the diketopiperazine (1, 5, 6). Amino acid resi- t. The t = 0 term is necessary to account for the racemization dues in diketopiperazines have been found to undergo rapid that already exists in the sample or for the racemization that racemization at both neutral and basic pH values (7-11). Ra- occurs when the sample processing procedure involves an cemization in aspartame solutions heated at neutral pH acid hydrolysis stem (17, 18). should thus take place primarily in the diketopiperazine de- Our aspartame experimental results, plotted in the form of composition product. Our results indicate that the racemiza- Eq. 1, are shown in Fig. 1. Several trends are apparent. With tion rates of aspartic acid and phenylalanine in the diketopi- the pure aspartame, aspartic acid at pH 4 is racemized more perazine at neutral pH are roughly similar. 3 A ,,ASP, B ,,' ~ AS 'ARTIC ACID

2 _ ,/ 0

t ,,,,Z,,,,,/,,---C..6 - /" o - 4 00 .4 -.2

a+1 01i'°/I 0 L 'c-| PHEI

HOURS FIG. 1. Racemization kinetics of aspartic acid and phenylalanine in solutions of pure aspartame (A) and Equal (B) heated at 100'C and pH 4 (o) and pH 6.8 (e). The slopes of the lines (obtained by a least squares fit of the data) are equal to 2 ki (see Eq. 1). The t = 0 point represents the amount of racemization that occurs during the HCI hydrolysis step. Chemistry: Boehm and Bada Proc. Natl. Acad. Sci. USA 81 (1984) 5265

Table 1. Racemization half-lives (hours) at 100'C for aspartic were about twice those of aspartic acid and phenylalanine in acid and phenylalanine in aspartame in comparison to the aspartame and were similar to those determined in the total rates in other systems solution. These results are in accord with the racemization hr mechanisms shown in Fig. 2. t½V2, In our experiments with Equal, the rates of racemization System pH Aspartic acid Phenylalanine of aspartic acid and phenylalanine at pH 4 were roughly simi- Aspartame 4.0 47 1200 lar to those determined by using the pure aspartame, where- 6.8 13 23 as at pH 6.8 the rates were slower. During the heating of the Equal* 4.0 43 500 Equal solutions, a marked brown coloration developed at pH 6.8 64 72 6.8. Although this brown color also formed at pH 4, the col- Free amino acids at 100C* 4.0 39 1200 oration was less pronounced than in the neutral pH experi- 6.8 630 1200 ments. This brown coloration is likely due to the formation Chicken muscle at 121Ct 4.7 630 of melanoidins-i.e., the condensation products of amino acids and peptides with carbohydrates (21-23). Since Equal The racemization half-lives are given by: tlf, = In 2/2ki. This is the contains cellulose and corn sweeteners, the reaction of time required to reach D/L = 0.33. *Commercially available product that besides aspartame also con- aspartame with these compounds accounts for the formation tains cellulose and corn sweeteners. of the brown color. Our results suggest that the rate of me- tTaken from Bada (17, 18). The phenylalanine rate at pH 4 was esti- lanoidin formation at neutral pH is greater than that at pH 4, mated to be the same as that at pH 7 based on measurements at which agrees with previous studies using free amino acids other temperatures (17, 18). and sugars (22). Once the aspartame is incorporated into the tTaken from Liardon and Hurrell (19). The muscle was heated in melanoidins, the formation of the diketopiperazine interme- sealed glass ampules in the presence of 15% water. diate would be inhibited (24). Thus, the formation of melan- At acidic pH values, the rate of aspartame decomposition oidins would act to reduce the racemization rates at pH 6.8, to its diketopiperazine is less than at neutral pH (1, 5, 6). which is what we observed. Also the rates of racemization of amino acid residues in diketopiperazines are slower at pH 4 in comparison to those in CONCLUSIONS the neutral pH region (7). Thus, racemization in the diketopi- At neutral pH, aspartic acid and phenylalanine in aspartame perazine formed from aspartame should not be significant at are extensively racemized in a day or so at 100°C. This race- mildly acidic pH values. Aspartame has its maximal stability mization takes place primarily in the diketopiperazine de- (1, 5, 6) at pH 3-5 and racemization should therefore occur composition product of aspartame and not in the sweetener primarily in aspartame (i.e., the dipeptide methyl ester) itself itself. In the presence of carbohydrates the racemization in this pH range. In our pH 4 experiments, the racemization rates at neutral pH are reduced (by a factor of -3-5), appar- rate of aspartic acid exceeds that of phenylalanine, which is ently due to the formation of melanoidin-like condensation consistent with the expected relative racemization rates at products. Our results indicate that the use of aspartame to acidic pH values of NH2-terminal aspartic acid and COOH- sweeten neutral pH foods or beverages that are then heated terminal phenylalanine (17, 18, 20). at elevated temperatures could result in the production of D- Our proposed mechanisms for the racemization of aspartic aspartic acid and D-phenylalanine. The quantity of D-aspar- acid and phenylalanine in aspartame at pH 4 and 6.8 are sum- tic acid produced from heating aspartame at elevated tem- marized in Fig. 2. To further substantiate these mechanisms, peratures and neutral pH could be equivalent to that which is we carried out racemization analyses of aspartame and its derived from the cooking of various foods (see refs. 19 and diketopiperazine in solutions heated for 24 hr. At pH 4, the 23 and Table 1). However, the amount of D-phenylalanine in racemization rates in aspartame and the diketopiperazine heated neutral pH mixtures that contain aspartame could were similar to those determined for the total heated aspar- substantially exceed that produced from cooking food. Be- tame solutions. At pH 6.8, the rates of racemization of aspar- cause phenylalanine is an essential amino acid and toxic to tic acid and phenylalanine in the diketopiperazine fraction phenylketonurics, its racemization could be of particular im- SCHEME SCHEME 2 0 0 (L)* II *(L) (L)* 1U *(L) 00C-CH2-CH -C -NH -CH -CH2-c6 HOOC-CH-CH-C-NH-CH-CH-() +NH COOMe +NH COOMe internal aminolysis followed by rapid I_ racemization at * centers 0 racemization HN CH2- at N-Terminal position HN *- H ( HH -OOC -C --- N H H2 0 11 (L,D)* I *(L) 0 HOOC-CH-CH-C-NH-CH-CH- ( 2 11 2 ( COOMe +NH3 FIG. 2. Mechanisms of aspartic acid and phenylalanmne racemization in aspartame at 100°C and pH 6.8 (scheme 1) and pH 4 (scheme 2). The symbol * denotes chiral centers. In scheme 2 the dipeptides are in equilibrium with the diketopiperazine but racemization takes place primarily in dipeptide methyl ester. The diketopiperazine can hydrolyze, producing not only the original dipeptide sequence Asp-Phe but also the inverted sequence Phe-Asp (7) and the various diasteromers of these dipeptides. 5266 Chemistry: Boehm and Bada Proc. Natl. Acad Sci. USA 81 (1984) portance. Aspartame loses its sweetness (1, 2) when heated 1. Harper, A. E. (1975) Sweeteners: Issues and Uncertainties at neutral pH, and is thus not recommended (2) as a sweeten- (National Academy of Sciences, Washington, DC), pp. 182- er under these conditions.t The potential formation of D- 188. considered as another important 2. Horwitz, D. L. & Bauer-Nehrling, J. K. (1983) J. Am. Diet. phenylalanine should be Assoc. 83, 142-146. reason for not using aspartame to sweeten neutral pH foods 3. Smith, R. J. (1981) Science 213, 986-987. that require cooking. 4. Beardsley, T. (1983) Nature (London) 305, 175. 'Most foods and beverages in which aspartame might be 5. Mazur, R. H. (1976) J. Toxicol. Environ. Health 2, 243-249. used as a sweetener have a pH in the range of 3-5. Our ex- 6. Furda, I., Malizia, P. D., Kolor, M. G. & Vernieri, P. J. (1975) periments indicate that in this pH range racemization-takes J. Agric. Food Chem. 23, 340-343. place in aspartame itself. However, the racemization rates 7. Steinberg, S. & Bada, J. L. (1981) Science 213, 544-545. are slow at 100TC, and only aspartic acid is prone to racemi- 8. Neuberger, A. (1948) Adv. Protein Chem. 4, 298-383. zation. Since the Arrhenius activation energies (Ea values) 9. Ott, H., Frey, A. J. & Hoffman, A. (1963) Tetrahedron 19, (17)' are 1675-1684. for free and peptide-bound amino acid racemization 10. Gund, P. & Veber, D. (1979) J. Am. Chem. Soc. 101, 1885- in the range of 25-30 kcal mold (1 cal = 4.184 J), it is appar- 1887. ent that at'room temperature aspartic acid and phenylalanine 11. Masters, P. M. & Friedman, M. (1980) Am. Chem. Soc. Symp. in aspartame would be stable with respect to racemization at Ser, 123, 165-194. mildly acidic pH 'values for periods in excess of 1000 years. 12. Hayashi, R. & Kameda, I. (1980) J. Food Sci. 45, 1430-1431. It is conceivable that some food and beverage components 13. Friedman, M., Zahnley, J. C. & Masters, P. M. (1981) J. Food could catalyze the racemization of aspartic acid and phenyl- Sci. 46, 127-131. alknine in aspartame. The nutritional and toxicological con- 14. Hoopes, E. A., Peltzer, E. T. & Bada, J. L. (1978) J. Chroma- sequences of D amino acids in the human diet are not well togr. Sci. X6, 556-560. further studies of 15. Manning, J. M. & Moore, S. (1968) J. Biol. Chem. 243, 5591- established. It would therefore seem that 5597. the racemization of aspartame in the presence of various 16. Bada, J. L. (1984) Methods Enzymol. 106, 98-115. food and beverage constituents are warranted to evaluate 17. Bada, J. L. (1982) Interdiscip. Sci. Rev. 7, 30-46. whether the use of aspartame as a'sugar substitute might not 18. Bada, J. L. (1984) in Chemistry and Biochemistry ofthe Amino result in an increased human consumption of D-aspartic acid Acids, ed. Barrett, G. C. (Chapman & Hall, London), pp. 399- and D-phenylalanine. 414. 19. Liardon, R. & Hurrell, R. F. (1983) J. Agric. Food Chem. 31, tThe box of Equal has an ambiguous statement that the product can 432-437. be used in recipes that "do not require the sweetener to be added 20. Kemp, D. S. (1979) in The Peptides: Analysis, Synthesis and before cooking." However, supermarkets sell products such as Biology, eds. Gross, E. & Meienhoffer, J. (Academic, New low-calorie hot chocolate that contain aspartame and that require York), Vol. 1, pp. 315-383. some exposure to elevated temperatures before consumption. 21. Hodge, J. E. (1953) J. Agric. Food Chem. 1, 928-943. 22. Hedges, J. I. (1978) Geochim. Cosmochim. Acta 42, 69-76. 23. Hurrell, R. F. & Finot, P.-A. (1983) Experientia Suppl. 44, We thank our colleagues at the University of California at San 135-156. Diego for numerous suggestions and N. C. Lee for reading several 24. Steinberg, S. & Bada, J. L. (1983) J. Org. Chem. 48, 2295- versions of this manuscript. 2298.