Electrophilic Properties of Patulin. N-Acetylcysteine ... - ACS Publications

phosphate buffer (pH 7.4) was studied by analytical HPLC/DAD, and most of the ... were confirmed by UV spectroscopy, formation of daughter products fr...
0 downloads 0 Views 296KB Size
Download PDF
Chem. Res. Toxicol. 2000, 13, 373-381

373

Electrophilic Properties of Patulin. N-Acetylcysteine and Glutathione Adducts Ralph Fliege and Manfred Metzler* Institute of Food Chemistry, University of Karlsruhe, P.O. Box 6980, 76128 Karlsruhe, Germany Received August 3, 1999

In our studies on the electrophilic properties of the mycotoxin patulin (PAT), we have now investigated the nonenzymatic reaction of PAT with the thiol-containing tripeptide glutathione and its metabolic degradation product N-acetyl-L-cysteine (NAC). Adduct formation in aqueous phosphate buffer (pH 7.4) was studied by analytical HPLC/DAD, and most of the products were isolated by preparative HPLC. Structure elucidation was carried out mainly by means of high-resolution NMR experiments and comparison of the data with those previously obtained for PAT adducts formed with simple model nucleophiles such as 4-bromothiophenol and 2-mercaptoethanol [Fliege, R., and Metzler, M. (2000) Chem. Res. Toxicol. 13, 363-372]. The assigned structures were confirmed by UV spectroscopy, formation of daughter products from isolated adducts, and partly FAB-MS. The reaction pathways of PAT with NAC were qualitatively the same as those previously observed for the aliphatic thiol model compound 2-mercaptoethanol. Due to the chiral nature of NAC and the new chiral center generated during the reaction with PAT, two diastereomers of each adduct were formed and observed in HPLC analysis. The major products formed in the reaction of PAT with GSH were of the same structural type as obtained with NAC. In addition, three cyclic adducts were formed with GSH, arising from the nucleophilic activity of the R-amino groups of the glutamic acid and the cysteine residue. In contrast, free cysteine yielded a markedly different adduct pattern, possibly due to the preferred formation of mixed thiol/amine-type adducts involving the R-amino group.

Introduction The mycotoxin patulin (PAT,1 Figure 1) is believed to exert its cytotoxic and chromosome-damaging effects mainly by forming covalent adducts with essential cellular thiols. One of the most likely cellular targets of PAT is the cysteine-containing tripeptide glutathione (GSH, γ-L-glutamyl-L-cysteinylglycine), which is present in millimolar concentrations in most cells. Indeed, a rapid depletion of GSH in living cells has been reported after treatment with PAT (1). GSH usually plays an important role in the detoxification of electrophilic xenobiotics or their metabolites, quenching them by enzymatic or nonenzymatic adduct formation involving its thiol group. Subsequent enzymatic degradation of the GSH adduct to the L-cysteine (CYS) and N-acetyl-L-cysteine (NAC) derivative facilitates the excretion of the xenobiotic. For certain xenobiotics, the reaction with GSH can lead to toxic metabolites. Although it has long been known that PAT is able to react with thiols, the chemical structures of such adducts are still unknown to date. We have therefore studied the nonenzymatic reaction of PAT with NAC, GSH, and CYS under in vitro conditions, and elucidated the chemical structures of the major adducts. Such basic knowledge * To whom correspondence should be addressed: Institute of Food Chemistry, University of Karlsruhe, P.O. Box 6980, 76128 Karlsruhe, Germany. Phone: +49-721-608-2132. Fax: +49-721-608-7255. Email: [email protected]. 1 Abbreviations: BTP, 4-bromothiophenol; CYS, L-cysteine; GSH, glutathione, γ-L-glutamyl-L-cysteinylglycine; ME, 2-mercaptoethanol; NAC, N-acetyl-L-cysteine; PAT, patulin, 4-hydroxy-4H-furo[3,2c]pyran2(6H)-one; TSP, 3-(trimethylsilyl)-2,2,3,3-propionic acid-d4, sodium salt.

Figure 1. Chemical structure of PAT.

of the fundamentals of PAT electrophilic reactivity toward small biological nucleophiles should help in future studies on the chemistry and biochemistry of this mycotoxin under cellular or in vivo conditions, as well as in the establishment of biomonitoring procedures.

Experimental Procedures Chemicals. PAT was a natural product of Penicillium expansum isolated and purified according to the procedure described previously (2). Its identity was checked by 1H and 13C NMR, IR, and UV spectroscopy. The substance was pure according to HPLC/DAD and GC/MS of the acetyl and trimethylsilyl derivatives obtained with acetic anhydride and N,O-bis(trimethylsilyl)acetamide, respectively. GSH (97% pure) and CYS (>98% pure) were purchased from Lancaster (Mu¨lheim/Main, Germany), and NAC (g99% pure) was from Fluka (Deisenhofen, Germany). All solvents and reagents for the preparation of buffers and HPLC eluents were of the highest commercial grade. High-Performance Liquid Chromatography. Two different HPLC systems were used for analytical and preparative HPLC. The analytical system consisted of two model 64 pumps (Knauer, Berlin, Germany), a model K45 gradient controller (ICI Kortec, Dingley, Australia), a type 7125 injector (Rheodyne, Cotati, CA) equipped with a 100 µL sample loop, and a model

10.1021/tx9901480 CCC: $19.00 © 2000 American Chemical Society Published on Web 04/15/2000

374

Chem. Res. Toxicol., Vol. 13, No. 5, 2000

SPD M6A photodiode array detector (Shimadzu, Kyoto, Japan) operated with Shimadzu CLASS M10A software version 1.4. The analytical column (250 mm × 4.6 mm) was filled with Polygosil60 5-C18 (Macherey-Nagel, Du¨ren, Germany). The preparative system comprised two model SD300 pumps (Rainin, Woburn, MA), a Rainin Dynamax UV-1 UV absorbance detector, a Rheodyne 7725i injector equipped with a 2 mL sample loop, and Rainin Dynamax Method Manager version 1.4 software for gradient programming and data acquisition. The preparative column (250 mm × 8 mm) was filled with Eurospher 100-C18, 5 µm (Knauer). Incubations and Separation of Adducts by HPLC. All incubations of PAT with NAC, GSH, and CYS were carried out in 1 M aqueous potassium phosphate buffer (pH 7.4). For studies on the time course of adduct formation monitored by HPLC, the PAT to thiol ratio varied from 1:1 to 1:25 with concentrations ranging from 5 to 20 mM for PAT, 20 to 250 mM for NAC and CYS, and 10 to 125 mM for GSH. The reactants were freshly dissolved in incubation buffer and the solutions mixed immediately to start the reaction, which was carried out at 20 °C for approximately 250 h. Aliquots of 100 µL were taken from the incubation mixture after various periods of time, acidified with 10 µL of formic acid, and analyzed by HPLC. The eluents were mixtures of methanol and water acidified with 1-10 mL of formic acid per liter. For the exact composition of the eluents and for gradient programs, see the legends of Figures 2, 5, and 9. The photodiode UV absorbance detector was operated in the full scan mode for the 200-400 nm wavelength range. In incubations prepared for the isolation of adducts, usually a 5-fold molar excess of thiol was used with concentrations of 20-40 mM PAT and 100-200 mM thiol. Incubation mixtures were kept at 20 or 60 °C and monitored by analytical HPLC until the desired adducts reached their maximum. Gradients and detection wavelengths for preparative HPLC (flow rate of 2-2.5 mL/min) were optimized for the respective separation problems, but in general resembled those applied in analytical HPLC (see above). The eluates were collected in ice-cooled test tubes and lyophilized either directly (GSH adducts) or after reduction of the methanol content in a nonheated evaporation centrifuge (NAC and CYS adducts). Characterization of Products. The chemical structures of the isolated products were elucidated using mainly 1H NMR, 13C NMR, and UV spectroscopy and FAB-MS. Comparison of the NMR and UV data with those obtained for the adducts of PAT with various model nucleophiles (see the Supporting Information of ref 3) proved to be very useful, as the proton NMR spectra sometimes contained regions with heavily superimposed signals or were complicated by the presence of diastereomers. In some cases, further NMR experiments (1H-1H COSY, 1H double resonance, and NOESY) were carried out. As the ion yield in FAB-MS was very poor even in an acidified matrix, no fragmentation could be detected and FAB-MS could only support the calculated molecular weight of the products by showing an [M + H]+ molecular ion peak in most cases. To reveal the pathways of adduct formation, some of the isolated adducts were re-incubated with thiol and analyzed for daughter product formation by analytical HPLC as described above. The instrumentation comprised a DRX 500 spectrometer (Bruker, Rheinstetten, Germany) for NMR measurements, a MAT 90 (Finnigan, San Jose, CA) mass spectrometer equipped with a cesium gun for FAB-MS, and an UVIKON 860 (Kontron, Zu¨rich, Switzerland) spectrophotometer for UV spectra. IR spectra of some of the adducts were recorded employing a Bruker IFS 88 Fourier transform spectrometer in the DRIFT mode; the samples were mixed with solid potassium bromide. NMR solvents were D2O (99.8 at. % D, Fluka, Neu-Ulm, Germany) for GSH and CYS adducts and D2O or methanol-d4 (99.8 at. % D, Aldrich, Milwaukee, WI) for NAC adducts. The 1H chemical shifts are expressed in parts per million with respect to the signal of sodium 3-(trimethylsilyl)-2,2,3,3-propionate-d4 (TSP, δ ) 0 ppm). 13C signals were referenced using dioxane as an external standard set to δ ) 67.8 ppm. To avoid

Fliege and Metzler

Figure 2. HPLC profiles and product peak assignment for the reaction of PAT (20 mM) with NAC (100 mM) in 1 M aqueous phosphate buffer (pH 7.4) at 20 °C. The column was a Polygosil 60 5-C18, 250 mm × 4.6 mm one; eluents were (A) 10% methanol and 5 mL/L formic acid and (B) 50% methanol and 5 mL/L formic acid, with a flow rate of 1 mL/min. The gradient was 100% A for the first 25 min, and then linearly changed to 100% B over the course of 40 min. The figure shows the UV 300 nm data trace of the DAD data set where all major adducts can be detected, even though some of them at a very poor response. Employing the Eurospher 100-C18, 5 µm, 250 mm × 8 mm preparative column, P1A and P1B could be separated, but P2A/B and P3A/B almost coeluted. P4A and P4B were more retarded and surrounded P5A+B. contamination of the GSH adducts with TSP, the downfield pseudo-singlet signal of glycine HR was used as a reference and set to 3.98 ppm, which was the value measured for GSH in D2O with respect to TSP as an internal standard. UV spectra were recorded in the double-beam mode in water versus water. The matrix for FAB-MS was usually glycerol; in some cases, a 1:10 mixture of trifluoroacetic acid and glycerol was used to increase ion yield. The extensive spectroscopic data are provided as Supporting Information.

Results Adduct Formation of PAT with NAC. When PAT was incubated with a 5-fold molar excess of NAC in aqueous phosphate buffer (pH 7.4) and the incubation mixture analyzed by HPLC after various periods of time, the HPLC profiles depicted in Figure 2 were obtained. All the reaction products could be detected at a wavelength of 300 nm, although some of them with an unfavorable response factor. When the ratio of NAC to PAT was lowered, the magnitudes of peaks 2A/B and 4A/B were increased in the HPLC profile whereas the magnitudes of peaks 5A+B, 6A/B, and 7A+B were decreased (data not shown). Conversely, increasing the NAC to PAT ratio gave rise to a more pronounced and faster formation of peaks 5A+B and 7A+B. As the pattern of reaction products depended strongly on the

N-Acetylcysteine and Glutathione Adducts of Patulin

Chem. Res. Toxicol., Vol. 13, No. 5, 2000 375

Scheme 1. Chemical Structures and Formation Sequences of the Major NAC and GSH Adducts of PATa

a Marked products (†) have not been isolated in sufficient amounts for NMR; their structures were derived from UV data, HPLC studies on daughter product formation, and comparison with the PAT adducts of model thiols, which were fully characterized (3).

molar ratio of NAC to PAT and on the incubation time, quantification of the adducts proved to be difficult. When the different absorption coefficients (see below) were taken into account, peaks 6A/B and 8A+B appeared to be the dominating products formed at a 2-10-fold excess of NAC and an incubation time of several hours, followed by peaks 5A+B, 7A+B, and 1A/B. To elucidate their chemical structures, the various adducts of PAT with NAC were separated and collected

by preparative HPLC, and subjected to spectroscopic analysis, in particular by NMR. From previous studies in our laboratory on the adducts of PAT with 4-bromothiophenol (BTP) and 2-mercaptoethanol (ME), the positions of the 13C and 1H signals of the PAT backbone in such adducts were known (3). Comparison of the NMR spectra of the NAC adducts with those of the BTP and ME adducts conclusively led to the reaction scheme presented in Scheme 1. This scheme, which also contains

376

Chem. Res. Toxicol., Vol. 13, No. 5, 2000

the respective GSH adducts of PAT (see below), closely resembles the product patterns found for BTP and ME. Four types of adducts were clearly formed in the reaction of PAT with NAC, viz. C-6- and C-2-thiol-substituted thioenolether ketones, ketohexanoic acids, and dihydropyranones. For a detailed discussion of the reaction sequence of PAT with thiols, see ref 3. Other types of adducts observed with BTP, e.g., the lactone type, and the initial adduct were not detected (3), which may well be explained by their short half-lives and their lack of strong UV chromophores. After a reaction time of 20 s, a short-lived broad HPLC peak became visible at 260 nm in the incubation of PAT with NAC (data not shown), which could possibly be attributed to some sort of initial adduct. In analogy to the aliphatic thiol model compound ME, C-2-thiol-substituted thioenolether ketone-type (P1A/B and P3A/B) and dihydropyranone-type (P4A/B) adducts can be observed, which could not be detected in incubations of PAT with BTP as an aromatic thiol compound. As in the case of ME, replacement of the C-5hydroxy group with another molecule of thiol was observed for the C-5-hydroxy C-6-thiol-substituted thioenolethers P5A+B, but does not appear to occur in the C-2thiol-substituted thioenolethers P1A/B. It is concluded from the structures of the adducts that the reaction of PAT with NAC follows the same pathways as outlined for the model thiol compounds in our previous publication (3). In contrast to the ME and BTP adducts, all NAC adducts of PAT contain multiple chiral carbon atoms. Whereas the R-carbons of the NAC moieties remain fixed in the L-geometry, the newly introduced chiral centers at backbone carbon C-6 or C-2 of PAT are formed in equal amounts of both stereoisomers. Therefore, two diastereomers of each adduct exist which can only in some cases be well separated by reversed-phase HPLC and are referred to by the suffix A and B of each adduct (see Figure 2 and Scheme 1). As both diastereomers exhibit very similar NMR spectroscopic data, the interpretation of spectra of diastereomeric mixtures is not severely affected. In the case of C-2-thiol-substituted thioenolether ketone adducts (P1A/B), the two diastereomers undergo a mutual conversion to form a 1:1 equilibrium within several hours in methanol-d4 solution, as observed during 1 H NMR measurements of the isolated pure diasteromers (data not shown). NOESY experiments were conducted with P1A and P7A+B to elucidate the E/Z-isomerism at the C-3dC-4 double bond of these thioenolether ketone-type adducts. Because the PAT backbone 4-H proton was found to interact with the 6-H proton(s) but not with the H-2 proton(s), an E*-geometry2 of the double bond is implied. As discussed previously for the ME adducts of PAT (3), the Z(*)- and E(*)-isomers of thioenolether ketone-type adducts can also be distinguished by the relative position of their 4-H proton signal in 1H NMR spectra which is slightly more downfield for E(*)-isomers (δ ≈ 8.2 and 8.1 ppm for C-6- and C-2-substituted products, respectively) than for Z(*)-isomers (δ ≈ 7.4 and 7.7 ppm, respectively). The thioenolether ketone-type adducts P1A and P7A+B as well as P5A+B, which is the C-5-hydroxy precursor 2 E*/Z* nomenclature of the double bond of C-2-thiol-substituted thioenolether ketones according to the relative positions of C-4 thiol and C-7 [for the sake of direct comparability with analogous C-6-thiolsubstituted thioenolether ketones; also, see ref 3].

Fliege and Metzler

Figure 3. UV absorption spectra and absorption coefficients for the major NAC and GSH adducts of PAT, measured in water vs water. The spectra that are depicted were obtained with the NAC adducts (except A and I), but were identical with those of the respective GSH adducts in parentheses: (A) PAT/GSH P10, (B) PAT/NAC P7 (PAT/GSH P9), (C) PAT/GSH P1 (≈PAT/GSH P5), (D) PAT/NAC P5 (PAT/GSH P6), (E) PAT/NAC P1 (PAT/ GSH P2), (F) PAT/NAC P4 (PAT/GSH P7), (G) PAT/NAC P6 (PAT/GSH P8), and (H) PAT/NAC P8. As in the ME model system (see ref 3), the spectra can be divided into three band shape types, typical for thioenolether ketones (type I), dihydropyranones (type II), and ketohexanoic acid derivatives (type III). The respective band shapes, maxima of absorbance, and absorption coefficients are identical to those of the structurally equivalent ME adducts. Table 1. Maxima of UV Absorbance and Assignment of Spectral Types for All NAC Adducts of PAT Detected by HPLC type I

type II

type III

peak no. λmax (nm) peak no. λmax (nm) peak no. λmax (nm) 1A/1B 2A/2B 3A/3B 5A+B 7A+B

302 323 315 306 308

4A/4B

272

6A+B 8A+B

294 294/231

of P7A+B (see Scheme 1), exhibit their 4-H proton NMR signals at δ ≈ 8.2 and 8.1 ppm for the C-6- and C-2-thiolsubstituted products, respectively. This corroborates the conclusion of an E(*)-configuration. The structural assignment for the NAC adducts is also in agreement with their UV spectra. Three types of UV spectra differing in the maximum of absorbance and shape of the bands were observed for the NAC adducts (Table 1 and Figure 3), in analogy to the ME and BTP adducts reported previously (3). Adducts P5A+B, P7A+B, and P1A/B exhibited “shorter-wavelength” type I UV spectra with λmax at 302-308 nm (Table 1), identical with those of the respective ME adducts. According to daughter product studies, P5A+B were formed from P2A/B and P1A/B from P3A/B (Scheme 1). From the longer-wavelength type I UV spectra for the precursors with λmax at 323 and 315 nm (Table 1), it was concluded that P2A/B and P3A/B represent the Z(*)-isomers of the C-6- and C-2thiol-substituted thioenolether ketones, respectively. As was found with BTP and ME, the Z-isomers were formed first due to kinetic or steric reasons and were later converted to the E(*)-isomers in a thiol-catalyzed isomerization process at the C-3dC-4 double bond (see ref 3). Because this isomerization occurred much faster with

N-Acetylcysteine and Glutathione Adducts of Patulin

Chem. Res. Toxicol., Vol. 13, No. 5, 2000 377 Table 2. Maxima of UV Absorbance and Assignment of Spectral Types for All GSH Adducts of PAT Detected by HPLC type I peak no. 1A/1B 2A/2B 3A/3B 4A/4B 5A/5B 6A+B 9A/9B 10A/10B

Figure 4. HPLC profiles and product peak assignment for the reaction of PAT (10 mM) with GSH (50 mM) in 1 M aqueous phosphate buffer (pH 7.4) at 20 °C. The column was a Polygosil 60 5-C18, 250 mm × 4.6 mm one; eluents were (A) water and 10 mL/L formic acid and (B) 15% methanol and 10 mL/L formic acid, with a flow rate of 1 mL/min. The gradient changed linearly from 5 to 10% B during the first 20 min, and then to 100% B over the course of the next 40 min. The figure shows the UV 300 nm data trace of the DAD data set where all major adducts can be detected, although some of them at a very poor response. Separation of P5-P7 (inset chromatogram) was achieved on the preparative column (Eurospher 100 C18, 5 µm, 250 mm × 8 mm). Eluents were (A) water and 2 mL/L formic acid and (B) 25% methanol and 2 mL/L formic acid, and the flow rate was 2.5 mL/min. The gradient changed linearly from 30 to 80% B within 30 min. Separation appears to be very dependent on the column material. With the preparative column, P6A+B could be well separated from P5A, P5B, P7A, and P7B. Separation of P1A and P1B was also possible, whereas P2A/B and P3A/B almost coeluted.

NAC and ME adducts than with BTP adducts, a Z-isomer of the C-5- and C-6-dithiol-substituted product (P7A+B) again could not be observed. In the final equilibrium, the E(*)-isomers predominated and were not converted to Z(*)-isomers to a significant extent when NAC was added. Adduct Formation with GSH. When PAT was incubated with a 5-fold molar excess of GSH in phosphate buffer at pH 7.4 for 43 h and aliquots of the reaction mixture were analyzed by HPLC after various periods of time, the chromatograms depicted in Figure 4 were obtained. All products could be detected at 300 nm. The UV spectra of the GSH adducts (Table 2) closely resembled and frequently matched the UV spectra of the NAC adducts, giving rise to the same three types based on the band shapes (Figure 3). A notable difference was the formation of the GSH adducts P1A/B, P5A/B, and P10A/B as additional products exhibiting type I spectra, which were not observed with NAC. Elevation of the GSH to PAT ratio favored the formation of peaks 6A+B, 9A/B, and 10A/B, whereas lowering the level of GSH decreased the magnitudes of those peaks and increased

type II

type III

λmax (nm) peak no. λmax (nm) peak no. λmax (nm) 307 301 323 316 307 306 308 337

7A/7B

272

8A+B

294

the magnitudes of peaks 3A/B and 4A/B. If the different UV absorption coefficients at the detector wavelength are taken into account, the major adducts formed with a 5-fold molar GSH excess and a 1 h incubation time at 20 °C are P6A+B, P7A/B, and P8A+B. The chemical structures of the GSH adducts, derived from the NMR and FAB-MS data of the isolated products, were to a large extent analogous to the structures of the major NAC adducts (Scheme 1). Like the NAC adducts, all GSH adducts appeared as two diastereomers (referred to as A and B) due to the newly generated chiral centers at C-6 or C-2 and the fixed chiral atoms within the GSH moiety. As a further similarity, NMR and UV data suggested a strong preference for the E(*)-isomers of the C-2- and C-6-thiol-substituted thioenolether ketone-type adducts over the respective Z(*)-isomers. Analytical support for the presence of Z(*)-isomers can be obtained from the short-lived HPLC peaks P3A/B and P4A/B, which exhibited type I UV spectra (Table 2) and were found to act as precursors of P6A+B and P2A/B, respectively. Apparently, these Z(*)-isomers again represent the initially formed species in this reaction, but are more rapidly transformed to the favored E(*)-isomers than the respective NAC adducts. In addition to these expected adduct types, three minor diastereomeric pairs of products were observed, viz. P1A/B, P5A/B, and P10A/B (Figure 4). According to their type I UV spectra, they can be assumed to represent thioenolether ketones, but the λmax for P10A/B (337 nm) did not match that of the established adduct type. A closer look at the 1H NMR spectra, which at first glance resembled those of the thioenolether ketone-type adducts, revealed two striking differences. (i) The singlet signal of the H-4 proton of P10A and -B at δ ) 7.55 ppm is typical neither for E(*)- nor for Z(*)-isomers of the known thioenolether ketone-type adducts. (ii) The integration of the usually isochronic protons of the GSH moieties indicated that one entire glutamic acid moiety of the apparently two molecules of GSH incorporated into the adduct was missing. This suggested that some kind of reaction leading to the loss of glutamic acid had taken place at the R-amino group of one of the cysteine moieties. The 13C NMR data also indicated the absence of one glutamic acid moiety. Moreover, a clear downfield shift of one cysteine R-carbon and an upfield shift of C-3 of the PAT backbone, which is located within the conjugated unsaturated system, were noted. The proposed structure of P10A/B and the mechanism of formation of these new thioenolether imine-type adducts of GSH are depicted in Scheme 2. The molecular weight of 621 of this structure is consistent with the FAB mass spectra exhibiting an (M + H)+ ion at m/z ) 622.

378

Chem. Res. Toxicol., Vol. 13, No. 5, 2000

Fliege and Metzler

Scheme 2. Proposed Structures and Formation of PAT/GSH P10A/Ba

a The reaction is hypothesized to involve a thiol-catalyzed conversion starting from PAT/GSH P6A/B or PAT/GSH P9A/B. The E/Zconformation at the C-3dC-4 double bond has not been elucidated.

The other two additional GSH adducts, P1A/B and P5A/B, also have spectra very similar to those of known thioenolether ketone adducts, but with some notable differences; the NMR signal of the 4-H protons at δ ≈ 7.7 ppm is typical for Z(*)-isomers of thioenolether ketones, but this structure would not be consistent with the UV spectrum with a λmax of 307 nm. Moreover, the NMR signal of the R-proton of one of the two glutamic acid moieties present in each adduct was markedly shifted downfield from 3.85 to 4.20 ppm. Furthermore, the signals of the respective diastereotopic β-protons were no longer isochronic but appeared as clearly distinguishable separate multiplets for this glutamic acid moiety. The 13C NMR data exhibited an upfield chemical shift of C-3 similar to that for P10A/B discussed above. Again, one of the R-carbons was clearly shifted downfield. Taken together with the proposed structure of peaks 10A/B, these findings suggest that in both P1A/B and P5A/B a reaction of one glutamic acid R-amino group with the C-7carbonyl function has taken place. As two GSH moieties were incorporated into these adducts, two possibilities exist for the formation of a cyclic adduct with an 11- or 12-membered ring system as depicted in Scheme 3. Because of the chirality at C-6, two diastereomers of each cyclic product are formed. The relevant NMR data of GSH adduct P1B are compared with those of P6A+B in Figure 5. Adduct Formation with Free CYS. Adduct formation of PAT with the free amino acid cysteine has previously been reported by Ciegler et al. (4, 5). In these studies, numerous reaction products were separated by

Figure 5. 1H and 13C NMR data of PAT/GSH P6 and P1, supporting the thioenolether imine hypothesis for PAT/GSH P1 involving one glutamine R-NH2 group.

TLC; however, no spectroscopic data were reported, and the structures of adducts proposed by Ciegler remain hypothetical. Various other authors have carried out toxicological studies employing uncharacterized “cysteine adduct mixtures” of PAT (6-10). When CYS was incubated with PAT in our laboratory and the mixture analyzed by HPLC/DAD, a multitude of products was formed even with a high molar excess of CYS (Figure 6). Some of the adducts had UV spectra similar to those of ME, NAC, and GSH adducts, whereas others exhibited quite different spectra. This implied that the R-amino group of CYS participated significantly in adduct formation in addition to the thiol group, as has already been hypothesized by Ciegler. As the cellular levels of free CYS are usually relatively low, we do not consider these adducts to be of greater toxicological importance and have only carried out some preliminary

N-Acetylcysteine and Glutathione Adducts of Patulin

Chem. Res. Toxicol., Vol. 13, No. 5, 2000 379

Scheme 3. Proposed Structures and Formation of PAT/GSH P1A/B and P5A/Ba

a Addition reactions of the R-amino groups of one of the GSH glutamic acid residues with the PAT backbone C-7 carbonyl group are assumed. The E/Z-conformation at the C-3dC-4 double bond has not been elucidated.

(Figure 6), was isolated and further characterized by NMR and FAB-MS. A proposed structure and mechanism of formation are depicted in Scheme 4. The sequence of processes leading from PAT to PAT/CYS P1 involves the R-amino group of CYS and many of the reactions previously observed for the formation of ketohexanoic acid adducts of PAT with mixtures of the model nucleophiles BTP and ethanolamine (3).

Discussion

Figure 6. HPLC profiles of the reaction of PAT (10 mM) with CYS (250 mM) in aqueous 1 M phosphate buffer (pH 7.4) at 20 °C. The column was a Polygosil 60 5-C18, 250 mm × 4.6 mm one; eluents were (A) water and 5 mL/L formic acid and (B) 25% methanol and 5 mL/L formic acid, with a flow rate of 1 mL/ min. The gradient was linearly changed from 100% A to 100% B within 45 min. The data traces represent the sum of the 270, 300, and 360 nm DAD signals, thus showing all products with UV absorption within the 240-400 nm range. Numbers above peaks indicate the respective absorbance maxima in nanometers. Only a few peaks fit the scheme of spectra types established for the ME, NAC, and GSH adducts, whereas the majority of peaks exhibit absorption band shapes and absorbance maxima at wavelengths not observed for known adducts.

studies to elucidate the structures. The PAT/CYS 1 adduct, which accumulated after prolonged reaction time

The study presented here continues our work on adducts of the electrophilic mycotoxin PAT with thiolcontaining nucleophiles. On the basis of the reaction pathways elucidated in a previous investigation with small model nucleophiles and the spectroscopic data collected for those “simple” adducts (3), it was now possible to establish the structures of the major adducts of PAT with the thiols NAC and GSH formed under in vitro conditions. Adduct formation with these biologically important nucleophiles proved to be quite analogous to the reactions with the aliphatic thiol ME; the major adducts of NAC and GSH can be divided into four structural classes, i.e., C-6-thiol-substituted thioenolether ketones, C-2-thiol-substituted thioenolether ketones, dihydropyranones, and ketohexanoic acid derivatives. All adducts gain one new chiral center at C-6 or C-2 of the PAT backbone during adduct formation. Because of the inherent and fixed chiral atoms of NAC or GSH, each of these adducts gives rise to two diastereomers. In addition to the types of adducts observed in the reaction of PAT with NAC or ME, three minor reaction products were formed with GSH (PAT/GSH 1, 5, 10, as pairs of two diastereomers each). Apparently, these additional products result from the reaction of PAT with

380

Chem. Res. Toxicol., Vol. 13, No. 5, 2000

Fliege and Metzler

Scheme 4. Structure and Proposed Formation of PAT/CYS 1a

a The pathway resembles the one suggested for the ketohexanoic acid derivative type (3), but involves aminolysis of the lactone ring by CYS-RNH2 instead of hydrolysis. In analogy to the equilibrium reaction observed for PAT/BTP/EA 1 and 2 (3), the resulting secondary amide is thought to attack the PAT backbone C-7 keto carbon, followed by water elimination to yield the final structure.

two molecules of GSH. 1H NMR spectral data suggest that in two of the adducts the free R-amino group of one glutamic acid residue has participated in a cyclization reaction, whereas in the third product, the R-amino group of cysteine is involved and one glutamic acid moiety is lost. 13C NMR data imply a thioenolether imine structure for these adducts instead of the thioenolether ketone structure of the corresponding major products. The data presented here should be useful for further analytical investigations on the fate of PAT in intact cells where adduct formation with GSH and possibly enzymatic conversion of these adducts to the respective mercapturic acid derivatives could play an important role. In addition to spontaneous adduct formation, processes mediated by GSH S-transferases can be expected, possibly leading to further types of adducts. On the other hand, possible effects of PAT on the conjugating enzymes have to be considered. Moreover, oxidative metabolism of PAT has to be taken into account, for which no experimental data are yet available. Therefore, the conditions in a cellular environment might well be more complex, and speculations about the in vivo relevance of our data should not be made at the current state of knowledge. For a proper monitoring of PAT adducts at toxicologically relevant low concentrations, in particular of the ketohexanoic acid type with low UV activity, new principles of detection will have to be established, e.g., cochromatography with radioactively labeled PAT derivatives or antibody-based analytical systems. A first indication that comparable reactions of PAT occur at micromolar concentrations is provided by studies showing a protein cross-linking potential of this mycotoxin in vitro (2). Knowledge of major in vitro adduct structures and their analytical characteristics presented in the current study could possibly initiate the synthesis of defined protein conjugates (e.g., of the dihydropyranone mono-

adduct type) for the generation of antibodies against those adducts or protein modification sites. Such antibodies should facilitate toxicological studies on the molecular mechanism of PAT toxicity and possibly lead to the establishment of biomonitoring procedures, e.g., based on PAT adducts in red blood cells, which are known to significantly retain [14C]PAT activity in rats (11). Moreover, they should be useful for the development of new antibody-based analytical methods for PAT. Until now, no antibody is available for the rapid and cheap analysis of PAT in foodstuffs, presumably due to the high reactivity of PAT against proteins. If antibodies could be generated against thiol adducts of PAT, these could be used to detect this mycotoxin in food samples after derivatization with a defined thiol.

Acknowledgment. We are greatly indepted to Dr. Herbert Roettele, Mr. Udo Tanger, and Mrs. Angelika Kernert from the spectroscopic laboratory of the Institute of Organic Chemistry, University of Karlsruhe, for their kind cooperation and expert measurements of NMR, FAB-MS, and IR spectra. This study was supported by a fellowship from the Fonds der Chemischen Industrie to R.F. and by a grant from the Deutsche Forschungsgemeinschaft (Me 574/14-1). Supporting Information Available: Spectroscopic data and NMR signal assignment for the isolated and characterized adducts. This material is available free of charge via the Internet at http://pubs.acs.org.

References (1) Barhoumi, R., and Burghardt, R. C. (1996) Kinetic analysis of the chronology of patulin- and gossypol-induced cytotoxicity in vitro. Fundam. Appl. Toxicol. 30, 290-297. (2) Fliege, R., and Metzler, M. (1999) The mycotoxin patulin induces intra- and intermolecular protein crosslinks in vitro involving cysteine, lysine, and histidine side chains, and R-amino groups. Chem.-Biol. Interact. 123, 85-103.

N-Acetylcysteine and Glutathione Adducts of Patulin (3) Fliege, R., and Metzler, M. (1999) Electrophilic properties of patulin. Adduct structures and reaction pathways with 4bromothiophenol and other model nucleophiles. Chem. Res. Toxicol. 13, 363-372. (4) Ciegler, A., Beckwith, A. C., and Jackson, L. K. (1976) Teratogenicity of patulin and patulin adducts formed with cysteine. Appl. Environ. Microbiol. 31, 664-667. (5) Ciegler, A. (1977) Patulin. In Mycotoxins in Human and Animal Health (Rodricks, J. V., Ed.) pp 609-624, Pathotox Publishers Inc., Park Forest South, IL. (6) Lindroth, S., and Wright, A. v. (1990) Detoxification of patulin by adduct formation with cysteine. J. Environ. Pathol. Toxicol. Oncol. 10, 254-259. (7) Lieu, F. Y., and Bullerman, L. B. (1978) Binding of patulin and penicillic acid to glutathione and cysteine and toxicity of the resulting adducts. Milchwissenschaft 33, 16-20.

Chem. Res. Toxicol., Vol. 13, No. 5, 2000 381 (8) Krivobok, S., Seigle-Murandi, F., Steiman, R., Benoit-Guyod, J.-L., and Bartoli, M.-H. (1994) Antitumoral activity of patulin and patulin-cysteine adducts. Pharmazie 49, 277-279. (9) Ciegler, A., Beckwith, A. C., and Jackson, L. K. (1976) Teratogenicity of patulin and patulin adducts formed with cysteine. Appl. Environ. Microbiol. 31, 664-667. (10) Cooray, R., Kiessling, K.-H., and Kiessling, K. L. (1982) The effects of patulin and patulin-cysteine mixtures on DNA synthesis and the frequency of sister-chromatid exchanges in human lymphocytes. Food Chem. Toxicol. 20, 893-898. (11) Dailey, R. E., Blaschka, A. M., and Brouwer, E. A. (1977) Absorption, distribution, and excretion of [14C]patulin by rats. J. Toxicol. Environ. Health 3, 479-489.

TX9901480