Formation of Vinylogous Compounds in Model Maillard Reaction

Sep 16, 2003 - File failed to load: https://cdn.mathjax.org/mathjax/contrib/a11y/accessibility-menu.js .... Apparently, intramolecular cyclization is ...
34 downloads 6 Views 403KB Size
Download PDF
1242

Chem. Res. Toxicol. 2003, 16, 1242-1250

Formation of Vinylogous Compounds in Model Maillard Reaction Systems Richard H. Stadler,* Ludovica Verzegnassi, Natalia Varga, Martin Grigorov, Alfred Studer, Sonja Riediker, and Benoit Schilter Nestle´ Research Center, Nestec Ltd., Vers-chez-les-Blanc, CH-1000 Lausanne 26, Switzerland Received May 9, 2003

The thermal degradation over temperature and time of selected amino acids (Asp, Gln, and Glu) in the presence of reducing sugars was investigated in low moisture model systems. Copyrolysis of glucose-Asp mixtures led to the release of acrylic acid, attaining >5 mmol/mol Asp at 230 °C after 5 min. Spurious amounts of 3-butenamide were detected upon heating Gln together with a carbonyl source. Apparently, intramolecular cyclization is favored to procure 2-pyrrolidinone, reaching levels >3 mmol/mol above 230 °C. 2-Pyrrolidinone was also formed in comparable amounts in pyrolyzed sugar-Glu mixtures, indicating that the Maillard reaction may be an important contributor to the formation of 2-pyrrolidinone in certain cooked foods. The chemical route to acrylic acid and 3-butenamide is probably analogous to that described for acrylamide recently. Evidence is also presented that acrylic acid may be an intermediate in the formation of acrylamide, and yields could be augmented by coincubation of fructoseAsp with certain amino acids such as Gln, reaching approximately 5% of the yield obtained by the Asn route. A computational study to determine the reactivity of the vinylogous products indicated a reduced ability of 3-butenamide as compared to acrylamide to form stable intermediates by Michael nucleophilic addition. Acrylamide and acrylic acid exhibited a similar theoretical reactivity potential toward nucleophiles. No information is as yet available on the occurrence of acrylic acid in cooked foods. Extensive toxicological evaluation indicates that acrylic acid is of no concern at the amounts to be expected in foods.

Introduction Since the discovery of relatively high levels of AA1 in many widely consumed foods (1), intensive studies have been launched to help understand how AA is formed. These investigations have revealed that the Maillard reaction plays a major role and that Asn directly provides the backbone of the AA molecule (2-6). Very shortly after the reports of AA in foods and the first ideas on its formation, hypotheses on the presence of other vinyl compounds formed in cooked foods by an analogous route were established, such as 3-But (7). A salient feature of the Maillard reaction under low moisture conditions is that the chemical pathway to the vinylogous products branches off at a very early stage (8). This step encompasses an intramolecular cyclization of the Schiff base to afford an oxazolidine-5-one intermediate, which provides conditions favorable for decarboxylation (9). Subsequent β-elimination of the decarboxylated Amadori product releases the vinyl compound (8). Should the pathway described by Yaylayan et al. (8) be generally applicable, then the composition of the food raw material (free amino acid pool and carbonyl source)

will be an important determinant in predicting the formation of the corresponding vinyl product in cooked or thermally processed food. Hence, Gln heated at moderate temperatures together with carbonyls or sugars could lead to the formation of 3-But and, similarly, Glu to the formation of vinylacetic (3-butenoic) acid, Asp to AC, etc. (Figure 1). In this study, model Maillard systems are employed to investigate the potential formation of vinyl compounds from the corresponding amino acid and sugar (carbonyl) reactants and to determine if the yields of such compounds in the model are comparable to those reported for AA (3). We chose those amino acids that are relatively abundant in their free form in different raw agricultural commodities, such as Gln, Glu, and Asp. In addition, the propensity of nucleophilic addition to the activated double bond of selected target compounds (AA, AC, and 3-But) is assessed by using Fukui function reactivity indices, derived through density functional quantum chemical calculations. Finally, the potential toxicological implications of the presence of related vinylogous reaction products in cooked foods are addressed.

Materials and Methods * To whom correspondence should be addressed. E-mail: [email protected]. 1 Abbreviations: AA, acrylamide; AC, acrylic acid; 3-But, 3-butenamide; Fruc, D-fructose, ESI, electrospray ionization; GABA, γ-aminobutyric acid; Gluc, D-glucose; 2-Pyr, 2-pyrrolidinone; HRMS, highresolution mass spectrometry; LC/MS, liquid chromatography/mass spectrometry; MRM, multiple reaction monitoring; NOEL, no observed effect level; NOAEL, no observed adverse effect level; SIM, selected ion monitoring.

Materials. AA (>99%), 2-Pyr, L-Asp, and AC were purchased from Sigma-Aldrich (Buchs, Switzerland). 3-But was custom synthesized by Toronto Research Chemicals (Toronto, Canada) (>99%). (1,2,3-13C3)AA (isotopic purity 99%), (1,2,3-13C3)AC (isotopic purity 99%), and (U-13C4)Asp (isotopic purity 98%) were purchased from Cambridge Isotope Laboratories (Andover, MA). Fruc and the amino acids L-Gln, L-Cys, L-Pro, L-Lys, and L-Glu

10.1021/tx034088g CCC: $25.00 © 2003 American Chemical Society Published on Web 09/16/2003

Maillard Reaction and Vinylogous Compounds

Chem. Res. Toxicol., Vol. 16, No. 10, 2003 1243 Table 1. Summary of Diagnostic Ions and MS Settings

Figure 1. Possible formation of vinylogous reaction products in the Maillard reaction. were purchased from Fluka (Buchs, Switzerland). Anhydrous Gluc was purchased from Merck (Darmstadt, Germany). Solid phase extraction cartridges (OASIS HLB 6 cm3, 0.2 g) were from Waters (Rupperswil, Switzerland). Multimode and NH2 cartridges (0.5 g) were from Isolute (Separtis, Germany). The filter units Spartan 13/0.2 RC were purchased from Schleicher & Schuell (Dassel, Germany). All other reagents were of analytical grade and were used without further purification. Caution: AA (CAS 79-06-1) is classified as toxic and may cause cancer. Wear suitable protective clothing, gloves, and eye/ face protection when handling this chemical. AC (CAS 79-10-7) is harmful by inhalation, in contact with skin, and if swallowed. Wear suitable protective clothing, gloves, and eye/face protection when handling this chemical. Standard Pyrolysis Procedures. The amino acid/sugar reactants (0.2 mmol unless otherwise stated) were heated in vacuum hydrolysis tubes (Pyrex vessels, 6 mL volume; all closed systems) on a temperature-controlled heating module (Brouwer). The tube insertions contained silicone oil to achieve an optimal contact with the tubes. A small volume of water (20 µL) was added to the reaction mixtures to promote physical interaction of the reactants. After the predefined heating period, the tubes were immediately placed on ice. The pyrolysates were suspended in 2.5 mL of water (for AA, 2-Pyr, and 3-But) or ammonium formate (5 mM, pH 7) for the determination of AC. All extracts for the determination of AA and AC were supplemented with 13C -labeled surrogate standard at 50 ng and 2.5 µg, respectively, 3 and the target analytes were extracted as described below. Extraction Procedures. 1. AC. The aqueous extract obtained after pyrolysis was vortexed (1 min) and sonified (5 min), and if required, an aliquot (1 mL) was centrifuged for 5 min at 9000 rpm (Eppendorf centrifuge). Under gravity-induced flow, 1 mL of the extract was loaded onto a NH2 cartridge (preconditioned with two bed volumes each of methanol and subsequently water). The cartridge was then washed with 0.5 mL of water (discarded), and AC was eluted with 1 mL of ammonium formate (0.2 M, pH 3.5, adjusted with concentrated formic acid). The effluent (1 mL) was loaded onto an OASIS cartridge and preconditioned with two bed volumes each of methanol and water. The column was eluted with 0.5 mL of a solution of ammonium formate (pH 3.5) 0.5 M:MeOH (20:80, v/v). The effluent was filtered (0.2 µm), and a portion (40 µL) was injected into the LC/MS system. 2. AA. The aqueous extract obtained after pyrolysis was treated as described for AC (vortexed, sonified, and centrifuged). Under gravity-induced flow, 0.5 mL of the clear extract was loaded onto an OASIS cartridge (preconditioned with two bed volumes each of methanol and water). The column was washed with 0.5 mL of water (discarded), and AA eluted with 0.5 mL of 20% (v/v) methanol in water. The effluent was filtered (0.2 µm), and a portion (50 µL) was injected into the LC/MS system. 3. 3-But. The procedure follows that described above for AA, except that extracts were not fortified with internal standard.

analyte

parent ion (m/z)

MRM (m/z)

AA 13C -AA 3 3-But 2-Pyr

72 [M + H]+ 75 [M + H]+ 86 [M + H]+ 86 [M + H]+

72 f 55; 72 f 54; 72 f 27 75 f 58; 75 f 29 86 f 69; 86 f 44; 86 f 41 86 f 69; 86 f 44; 86 f 41

collision cone energy voltage (eV) (V) 11; 20; 20 11; 20 11; 20; 20 11; 23; 23

22 22 22 41

4. 2-Pyr. The aqueous extract obtained after pyrolysis was treated as described for AC (vortexed, sonified, and centrifuged). Under gravity-induced flow, 0.5 mL of the clear extract was loaded onto a Multimode cartridge, preconditioned with two bed volumes each of methanol and then water. The column was washed with 0.5 mL of water (discarded), and 2-Pyr eluted with 30% (v/v) methanol in water (0.5 mL). The effluent was collected and diluted 9 + 1 with water (v/v) and filtered (0.2 µm), and a portion (50 µL) was injected into the LC/MS system. LC-MS Conditions. 1. AA, 3-But, and 2-Pyr. Mass spectrometry measurements were performed using a Waters separation module Alliance 2690 coupled with a Quattro LC mass spectrometer (Micromass, Manchester, U.K.). Analytical separation was achieved as described by Riediker and Stadler (10) employing a Shodex RSpak DE-613 HPLC column (polymethacrylate gel, 150 mm × 6 mm i.d., Showa Denko K. K., Japan). The elution mode was isocratic, using a mixture of methanol and water (40:60, v/v) containing 0.06% (v/v) of formic acid as LC eluent. The initial flow rate was set at 0.75 mL/min and reduced by postcolumn splitting after the LC column to 0.25 mL/min. The column temperature was maintained at 40 °C with a column heater. Typical retention times (tR, min) under these conditions for 2-Pyr, AA, and 3-But were 4.9, 5.1, and 5.9, respectively. The analytes were detected by MRM in the positive ESI mode. Two or three different fragment ion transitions were monitored for each analyte and internal standard and are shown in Table 1. The capillary voltage was set to 3.25 kV, and the ion energy was set to 0.9 and 1.0 V for the first and second quadrupole, respectively. The desolvation and source block temperatures were set at 350 and 100 °C, respectively. Nitrogen was used as the nebulizer (100 L/h) and desolvation gas (600 L/h). Argon was used as the collision gas at a pressure of 1.7 mTorr (2.3 mbar). 2. AC. Mass spectrometry measurements were performed using an Agilent 1100 HPLC (Agilent Technologies, Waldbronn, Germany) coupled with a Quantum Finnigan mass spectrometer (Spectronex AG, Basel CH). Analytical separation was achieved by using a Hamilton PRP-X100 5 µM HPLC column (anion exchange, 50 mm × 2.1 mm i.d., Hamilton, Reno, NV). Two solvents were used for the elution of AC: (A) 10 mM ammonium formate (pH adjusted to 8 with diluted aqueous ammonia) and (B) 100% methanol. The elution was isocratic from 0 to 2 min (25% of B). A linear gradient was then used increasing to 40% B at 7 min and reverting back to 25% B at 10 min. The elution continued isocratic until 15 min with 25% B. The typical retention time under these conditions for AC was 6.3 min. The MS settings were as follows: dwell time, 0.5 s; electrospray capillary, 3.3 kV; tube lens offset, -120 V; capillary temperature, 325 °C. The sheath and auxiliary gas flows were set to 40 and 5 psi, respectively. For AC and its isotopic surrogate, only one mass was monitored in the SIM mode, m/z 71 and m/z 74. HRMS. Accurate mass measurements for AC were performed on a Finnigan MAT8430 (Finnigan, Bremen, Germany) mass spectrometer. Samples were introduced via the direct inlet probe, and peak matching was performed at a resolution of ca. 4000 (10% valley definition) with perfluorokerosene as the reference compound. To confirm the identity of 2-Pyr, accurate mass measurements were performed by ESI coupled to a highresolution time-of-flight instrument, similar to a technique described earlier (11). The sample was mixed with 50 µL of calibrant solution (1 mg/mL NaI, 0.05 mg/mL CsI in 2-propanol), and HRMS were then acquired by infusion of the sample solution into the ion source of a Micromass QToF-2 instrument (Micromass). Data acquisition and data evaluation were per-

1244 Chem. Res. Toxicol., Vol. 16, No. 10, 2003 formed using the Micromass MassLynx 4.0 software. The temperature of the instrument’s ion source was set to 90-110 °C, the electrospray voltage was set to 3.5 kV, and the cone gas (nitrogen) was set to a flow rate of 60 L/min. About 20 scans were averaged at different cone voltages (2 min total acquisition time, ca. 1 scan/s), and accurate mass data were obtained using a suitable lockmass from the calibrant. The corresponding elemental compositions were calculated as described (11). Computational Studies. Quantification of chemical reactivity involves the evaluation of the responses of the electrons of a molecule to a given external perturbation. It is possible to describe a chemical reaction in terms of electronic reactivity indices of the reagents. These indices account for the responses of the electrons to a local or global change in the total electron number of the molecule. According to the original results of Parr and Yang (12), the chemical reactivity of a given molecule is determined by three basic behaviors: reactivity toward electrophiles, defined by the local Fukui function f +(r), toward nucleophiles defined by the function f -(r), and reactivity toward radicals, defined by the Fukui function f 0(r). Within the frozen orbital approximation, i.e., by assuming that molecular orbitals are not deformed upon electron transfers, these three Fukui functions are defined by the following equations:

f +(r) ∼ FHOMO(r) f -(r) ∼ FLUMO(r) f 0(r) ∼ 1/2[FHOMO(r) + FLUMO(r)] where FHOMO(r) and FLUMO(r) are the electron densities of the highest occupied molecular orbital and the lowest unoccupied molecular orbital, respectively. Fukui functions reactivity indices were computed for the lowest energy molecular geometries adopted by AC, AA, and 3-But by using the DeFT software. Geometry optimizations at the local density level of approximation were carried out by employing the Vosko-Wilk-Nussair functional for correlation. Extended (7111/411/1*) basis sets for the carbon, nitrogen, and oxygen atoms and (41/1*) basis sets for the hydrogen atoms have been used. Again, extended (5,2;5,2) auxiliary basis sets were used for the carbon, nitrogen, and oxygen atoms, as well as (5,1;5,1) auxiliary basis sets for the hydrogen atoms. A 64 points nonrandom fine integration grid was used. Only two starting geometries were considered for AC and AA, originating from free rotation around the Ccarbonyl-Csp2 bond. In the case of 3-But, which exhibits much higher molecular flexibility, we conducted a 200 ps molecular dynamics run at 300 K in a water box following an equilibration period of 20 ps. Conformers were collected at every picosecond along the trajectory and clustered on the basis of root mean squared deviations taking in account heavy atoms only. All molecular dynamics computations were executed within the Sybyl molecular modeling software package by using the Tripos force field. The geometry of the lowest energy conformation derived for 3-But is in good agreement with the one previously published (13). Quantification of AA, 2-Pyr, and 3-But. Standards were prepared in Millipore-grade water from a stock solution (0.01 mg/mL) and stored in the refrigerator for 2-4 weeks. Calibration curves (five point) were established in the concentration range from 10 to 1000 pg/µL analytes, containing for AA a fixed amount of internal standard (20 pg/µL). 2-Pyr and 3-But were quantified using an external calibration curve, choosing the transitions m/z 86 f 69 and m/z 86 f 44, respectively, for quantitation. For additional certainty of the analyte, ion ratios of 69/44 (2-Pyr ) 3.2; 3-But ) 0.3) and 69/41 (2-Pyr ) 2; 3-But ) 1.4) were compared to the standard compounds (acceptance if within 10% of the average value obtained on standard compounds, n ) 20). Typical recoveries for 2-Pyr were in the range of 80-96% at a spiking level of 200 ng/mL in water, and because no internal standards were available, data for 3-But and 2-Pyr are not corrected for recovery loss. Both analytes

Stadler et al.

Figure 2. Typical excerpt of a SIM chromatogram of a pyrolyzed Fruc-Asp mixture showing traces (top) for AC and (bottom) for (13C3)AC (tR of AC ) 6.3 min). showed comparable sensitivity in our analytical assay, and 2530 pg could be quantified on-column (considering three characteristic transitions for each analyte). For 2-Pyr, potential ion suppression by coelution of matrix constituents was largely avoided due to the strong dilution (10-fold) of the extracts prior to LC/MS analysis. All data evaluation were done using the MassLynx software, and in the case of AA, data were normalized to the area response analyte to internal standard. For all standard curves, r2 > 0.99. Samples with analyte levels not within the dynamic range of the standard curve were diluted as to achieve an acceptable analyte range. Quantification of AC. Standards were prepared in Millipore-grade water from a stock solution (1 mg/mL) and stored in the refrigerator for 4 weeks. The calibration curve (five point) was established in the concentration range from 50 to 10 000 pg/µL, containing a fixed amount of isotopically labeled internal standard (1 ng/µL). Using this assay, an on-column detection limit of 0.5 ng could be achieved for AC. All data evaluation was done using Xcalibur software, and the area response of AC was normalized to that of the corresponding surrogate.

Results Formation of AC in Sugar-Asp Copyrolysates. AC was determined in Maillard reaction mixtures using isotope dilution LC/MS, selectively monitoring in negative ESI mode m/z 71 [M - H]- and m/z 74 for the analyte and isotope-substituted internal standard, respectively. A typical SIM chromatogram of an extract of a FrucAsp mixture heated at 230 °C for 20 min is illustrated in Figure 2. Under our MS conditions, AC could not be fragmented; thus, confirmation of the compound was obtained by independent HRMS analysis of a typical pyrolysate (found, 72.0248 Da; calcd, 72.0211 Da). Further evidence that Asp is the carbon source of AC was obtained by heating (230 °C, 5 min) an equimolar mixture of (U-13C4)Asp and Fruc. Subsequent analysis of the extract by MS showed >98% incorporation of the 13Clabel into CH2dCHCOOH. The condensation of equimolar amounts of Gluc with Asp over a temperature range of 150-240 °C led to a yield (>5 mmol/mol amino acid at 230 °C and after 5 min) of AC (Figure 3A), comparable to that of AA from an analogous reaction with Asn using the same fundamental model (3). Mild pyrolysis of the amino acid alone did not yield AC. Substitution of Gluc for Fruc afforded a significantly higher yield of the vinyl product, especially at temperatures >180 °C, which is coherent with many of our and other (4, 14, 15) observations that Fruc appears to be a more effective reactant. In this model system, a maximum yield of AC from a Fruc-Asp copyrolysate in a closed system was measured after 5 min at a constant temperature of 230 °C and declined slightly

Maillard Reaction and Vinylogous Compounds

Chem. Res. Toxicol., Vol. 16, No. 10, 2003 1245 Table 2. Formation of AA (µmol/mol Asp) in Fruc-Asp Pyrolysates (0.2 mmol of Reactants, 200 °C, 10 min) Supplemented with Other Amino Acids at Different Amountsa amino acid supplemented control (only Fruc-Asp) Gln Fruc-Gln onlyb Lys Fruc-Lys onlyb Pro Fruc-Pro onlyb Cys Fruc-Cys onlyb

amount (mmol) 0.2 0.4 1.0 0.2 0.4 1.0 0.2 0.4 1.0 0.2 0.4 1.0

AA ( SD 2.1 ( 0.7 47.3 ( 5.6 70.5 ( 13.3 74.5 ( 11.2 14.8 ( 1.2 7.6 ( 1.5 7.4 ( 1.3 4.3 ( 0.2 ND 4.4 ( 1.9 4.4 ( 1.3 2.9 ( 0.7 ND 1.2 ( 0.8 0.17 ( 0.14 ND 1.13 ( 0.2

a All entries are averages ( SD of at least four independent determinations. b Incubations without Asp, each 0.2 mmol of reactants; ND, not detected.

Figure 3. Formation of AC from sugar-Asp pyrolysates (2, Gluc-Asp; 0, Fruc-Asp) as a function of (A) temperature (pyrolysis time 5 min) and (B) time (230 °C). All entries are averages ( SD of n ) 4 independent determinations.

over the following 15 min (Figure 3B). However, even with an internal standard, a larger variance in the data was observed at temperatures >220 °C, probably attributable to rapid formation-decomposition of AC as already reported for AA (16). Formation of AA in Sugar-Asp/Amino Acid Cocktails. The intermediacy of AC in the formation of AA has already been proposed (2, 4), but so far, no concrete evidence on the formation of AC in a Maillard model has been provided. We were prompted to measure AA in a number of amino acid mixtures after the finding that AC is released in sugar-Asp condensates at moderate temperatures. As shown in Table 2, incubation (200 °C, 10 min) of equimolar concentrations of Fruc-Asp led to low but detectable levels of AA, albeit approximately 1000fold below the levels recorded with Fruc-Asn mixtures (3). Heating equimolar amounts of Fruc-Gln-Asp clearly enhanced (22-fold) the formation of AA as compared to incubations without Gln. Incubation of Fruc-Gln as a control yielded only a fraction of the AA (24.5%), indicating that the intermediacy of AC is important in the reaction. The amount of AA released could be enhanced by augmenting the molar ratio of Gln/Asp in the mixtures (2-fold) but did not increase further, probably due to competitive reactions of the amines for the carbonyl source. An important aspect to consider in model test systems with individual amino acids is their chemical purity (i.e., degree of contamination with foreign amino acids), a valid point that was raised recently (6). Therefore, 1H NMR measurements (600 MHz) on commercially available Gln were conducted to assess the degree of potential contamination with Asn. Asn could not be detected in the Gln batch that was used in the experiments, with a

sensitivity estimated at 4 mmol/mol amino acid)

1246 Chem. Res. Toxicol., Vol. 16, No. 10, 2003

Stadler et al.

Figure 5. Formation of 2-Pyr from (A) sugar-Gln (2, GlucGln; 0, Fruc-Gln) and (B) sugar-Glu (2, Gluc-Glu; 0, FrucGlu) over temperature (incubation time, 5 min). All entries are averages ( SD of n ) 3 independent determinations.

Figure 4. Product ion spectrum of (A) 3-But and (B) 2-Pyr, both dissolved in 0.01% formic acid in water/methanol (6:4, v/v). Cone voltage/collision energy (V/eV): 22/20 (3-But) and 41/23 (2-Pyr).

at 240 °C after a heating period of 5 min. Substituting Fruc for Gluc showed a relatively lower turnover rate. This finding prompted additional experiments with sugar-Glu mixtures, which after numerous attempts failed to procure detectable amounts of 3-butenoic acid. Thus, we could show also in these experiments that 2-Pyr is an important reaction product, attaining similar yield as the Gln mixtures (Figure 5B). Gln and Glu both follow similar temperature kinetics, supporting a common mechanism to the formation of 2-Pyr (see Discussion). Computational Studies of Vinyl Compounds. The Fukui functions f +(r) (electrophiles), f -(r) (nucleophiles),

and f 0(r) (radicals), determining the three basic reactivity patterns of AA, AC, and 3-But, were computed by density functional theoretical methods and are illustrated in Figure 6. Fukui functions were computed for the energetically preferred conformations of AA, AC, and 3-But. Reactivity toward nucleophiles was further investigated, as this is expected to determine most of the toxic potential. Predictions about the reactivity could be formulated in the assumption that chemical reactions occur by superposition of the positive lobes in the reactants Fukui functions. On the basis of this assumption, it is expected that the largest isodensity lobe in the f -(r) Fukui function will be the site of the preferred nucleophilic attack. A similar conceptual framework has been applied previously to derive quantitative correlations between the cytotoxic activities of a series of sugar derivatives with electrophilic indices, defined by the isovalue surfaces of the interaction energy with an incoming model nucleophile (H-). Different to our study, the indices were calculated from extended Hu¨ckel wave functions (18). In view of these results, we further propose to use the volumes of the isodensity lobes of the f -(r) Fukui function as quantitative measures of the nucleophilic reactivities of the three compounds under scrutiny. The volumes of the isodensity lobes at 0.01 atomic units were approximated by the volumes of the respective convex hulls. The convex hull of a set of points is the smallest convex set that includes the points, where in two dimensions the convex hull reduces to a convex polygon. We used the software qhull to compute the volumes of the convex hulls in the three-dimensional space (19), which

Maillard Reaction and Vinylogous Compounds

Chem. Res. Toxicol., Vol. 16, No. 10, 2003 1247

Figure 6. Positive lobes of electron density of the Fukui functions f -(r), f +(r), and f 0(r) governing the reactivity behavior of AC, AA, and 3-But toward electrophiles, nucleophiles, and radicals, respectively. Lobes are determined as the electron isodensity surfaces at 0.01 atomic units. Sites of preferred attack are indicated by arrows.

were found to relate as 1.25:0.78:1.10 for AA:AC:3-But. The same principle was applied for deducing the preferred sites for electrophilic and radical attacks, and these are indicated by the arrows in Figure 6.

Discussion A key event leading to the release of vinyl compounds such as AA is the very early thermal decarboxylation of the Schiff base through a oxazolidin-5-one intermediate (8, 9). A competitive reactionsand thus favored to avoid the formation of double bond reaction productssis the rearrangement of the imine to the corresponding Amadori product, which can further be degraded in multiple steps to a plethora of compounds via the classical Maillard route. Using a model system, we demonstrate that other pathways are feasible in the formation of AA, not requiring Asn as a reactant but rather compounds that can generate AC. This intermediate could react with ammonia by amino dehydroxylation (20) to furnish AA. Hence, the formation of AC is a determinative step in the process, and AC can be released in the Maillard reaction by an analogous route from carbonyl-Asp interactions. Supplementation of a Fruc-Asp reaction mix with additional amino acids known to deamidate at relatively low temperatures (110 °C) such as Gln (17) may provide the important ammonia source. AC could, however, also originate from acrolein, by a multistep process from the thermal degradation of lipids/glycerol (21), carbohydrates (22), amino acids (23), or Maillard products (24). The potential role of acrolein in the generation of AC warrants additional study and will not be elaborated further in this paper. On the basis of the results obtained in the model systems, both AC and AA could be formed in foods with an abundance of free Asp and the availability of sugars, ammonia being generated by the decomposition of amino

acids present either in the free or in the bound form. AC has to our knowledge never been considered in food as a processing-derived compound; thus, most of the data available in the literature is related to (poly)acrylates used in food packaging materials (25). However, at this stage, it is not possible to predict whether the AC pathway is only of marginal importance or may be a favored mechanism in any cooked food. Inspection of different foods known to release AA but with low levels of free Asn may provide a clearer picture. Furthermore, AC is freed at relatively high temperatures (optimum > 220 °C), suggesting that thermal energy barriers may be higher for the reaction to proceed. Following similar mechanistic principles, thermal degradation of carbonyl-Gln and carbonyl-Glu mixtures could procure 3-But and 3-butenoic acid, respectively. As shown in our experiments, these reactions are apparently not favored and both copyrolysates release significant amounts of 2-Pyr as a major degradation product. Thermally catalyzed intramolecular cyclization probably of the decarboxylated glycosylamine (Figure 7) followed by heterolytic cleavage could release 2-Pyr, the sugarfacilitating ring closure, and the scission of the C-N bond. Comparable temperature kinetics of formation of 2-Pyr from both Gln and Glu copyrolyzed with hexose sugars support a common mechanism (see Figure 5). Under our experimental conditions, 3-But was only present in spurious amounts, indicating less importance in cooked foods. Furthermore, 3-But is relatively stable under moderate temperatures and does not cyclize to 2-Pyr. Therefore, the formation of corresponding vinylogous compounds in foods is not clear-cut (see Figure 1) and must be studied on a case-by-case basis, taking into account the availability of free amino acids and sugars in the raw agricultural commodity and thermal-processing conditions.

1248 Chem. Res. Toxicol., Vol. 16, No. 10, 2003

Stadler et al.

Figure 7. Proposed mechanism for the formation of 2-Pyr in the Maillard reaction from sugar-Gln or sugar-Glu copyrolysates.

Gln and Glu are abundant amino acids present in a free form in many different foods such as wheat (26, 27), soybean (28), almonds (28), cocoa (29), and coffee (30). Hence, the Maillard reaction may contribute significantly to the formation of 2-Pyr levels in cooked and roasted foods. The available data confirm this hypothesis (31, 32). Roasted materials contain significant amounts of 2-Pyr, with levels up to 77, 50, and 32 mg/kg in cocoa powders, coffee beans, and coffee substitutes, respectively (31). Lower concentrations were observed in roasted products as consumed, such as 0.5-1 mg/kg in coffee beverage preparations, 2 mg/kg in cocoa beverages, 2 mg/kg in chocolate, and 9 mg/kg in bitter chocolate (31). Other sources of 2-Pyr are also possible. The natural content of 2-Pyr in food plants has been reported to be in the range of 0.1-2.2 mg/kg (33) while high levels have been found in dried vegetables (e.g., 48 mg/kg in dried tomatoes) (31). A proposed source of 2-Pyr in dried vegetables is GABA, which under dehydrating conditions could ring close to the lactam. Monosodium glutamate has also been proposed to be a direct precursor of 2-Pyr, which upon thermal treatment leads to decarboxylation and ring closure (34). However, the thermolysis of Gln and Glu alone (200 °C, 5 min) without a carbonyl source yielded only trace amounts of 2-Pyr (