Patulin is a mycotoxin produced by molds of the genera

Transcription

1 104 ROACH ET AL.: JOURNAL OF AOAC INTERNATIONAL VOL. 83, NO. 1, 2000 FOOD CHEMICAL CONTAMINANTS Capillary Gas Chromatography/Mass Spectrometry with Chemical Ionization and Negative Ion Detection for Confirmation of Identity of Patulin in Apple Juice ROACH ET AL.: JOURNAL OF AOAC INTERNATIONAL VOL. 83, NO. 1, 2000 JOHN A.G. ROACH, KEVIN D. WHITE, MARY W. TRUCKSESS, and FREDERICK S. THOMAS U.S. Food and Drug Administration, Center for Food Safety and Applied Nutrition, 200 C St, SW, Washington, DC Gas chromatography/mass spectrometry (GC/MS) with negative ion chemical ionization permits detection of underivatized patulin in apple juice extracts while minimizing co-extractive responses. The technique has been used with a variety of capillary columns in quadrupole, ion trap, and magnetic sector GC/MS instruments to confirm presumptive findings of patulin in apple juice at concentrations ranging from 68 to 3700 g/l. The demonstrated ability to use any of these 3 mass spectrometers and several capillary columns to confirm the identity of patulin are significant strengths of the technique. Patulin is a mycotoxin produced by molds of the genera Aspergillus and Penicillium. It can be found in moldy fruit, vegetables, and cereals. The level of contamination correlates with the degree of spoilage, and the patulin is confined to the spoiled tissues (1). Use of moldy produce in the production of fruit juices and processed foods may result in the presence of patulin in the finished food products. The Joint Expert Committee on Food Additives of the World Health Organization and the Food and Agriculture Organization has evaluated the hazard posed by patulin based on the determination of a no-observed effect level (NOEL) in long-term toxicological studies (2). A safety factor derived by dividing the NOEL by 100 fixed the maximum tolerable daily intake level of patulin in 1995 at 0.4 µg/kg body weight. Patulin is soluble in water and polar organic solvents. It is not destroyed by heat. It is stable at an acid ph, but is destroyed by base or fermentation. Penicillium expansium is primarily responsible for rot in apples and other fruit. When apples infected by this organism are used to produce fruit juice, the final product is likely to contain patulin. Available worldwide analytical results for products that may be contaminated with patulin show that at least 30% of apple juices may be contaminated at levels >50 µg/l. The consumption data for apple juice indicate that certain consumers imbibe as much as 250 ml apple juice per day (1). Received March 16, Accepted by AP August 30, In the United States, background exposure data and toxicological data are being obtained and evaluated for patulin. Recent studies have demonstrated that patulin has immunosuppressive effects that may be of concern in certain compromised populations (3). It also has antibiotic activity and has caused dermal and gastric irritation in human subjects (4). The European regulation of patulin at 50 µg/l has prompted American industry to petition the U.S. Food and Drug Administration (FDA) to set 50 µg/l as an action level to keep products rejected by other countries from being exported to the United States. Action levels are intended to represent the best guidance available on levels that FDA considers to be of regulatory interest, and they should enhance the safety of the food supply when used and adhered to by the food industry (5). The FDA has been monitoring apple juice for patulin since Patulin levels found are generally <50 µg/l. At present, findings of patulin are being handled on a case-by-case basis. The levels of patulin that have been confirmed by mass spectrometry (MS) in our laboratory in support of presumptive findings by liquid chromatography (LC) have ranged from 68 to 3700 µg/l (ppb). The levels were quantitated by LC. The results of a collaborative study of the LC method for the determination of patulin in apple juice were recently reported (6). The confirmatory assay used in our laboratory was gas chromatography/chemical ionization (GC/CI) MS of underivatized patulin with negative ion detection. Omission of the formation of a patulin derivative (7 9) saves time and does not detract from the purpose of the analyses. The rationale for this approach and representative data are presented here. Experimental Principle Patulin was extracted with ethyl acetate and cleaned up by extraction with sodium carbonate solution. The extract was dried with anhydrous sodium sulfate. Patulin is determined by reversed-phase LC with UV detection in aqueous acidic solution after evaporation of the ethyl acetate (6). Presumptive patulin findings were confirmed by capillary GC/MS.

2 ROACH ET AL.: JOURNAL OF AOAC INTERNATIONAL VOL. 83, NO. 1, Apparatus and Chromatographic Conditions (a) High-resolution magnetic sector GC/MS instrument. Micromass (Manchester, UK) Autospec Q mass spectrometer coupled to a Hewlett Packard 5890 Series II gas chromatograph (Wilmington, DE). The capillary column was 30 m 0.32 mm id with a 0.25 µm film of bonded phase 5% phenyl/methyl polysiloxane. Helium carrier gas head pressure was set at 14 psi with a split flow of 100 ml/min. Injections were splitless via the split/splitless injector. Temperatures were 200 C injector, 265 C transfer lines, and 250 C ion source. The GC oven was held at 55 C for 1 min after injection with septum purge OFF, and then heated with septum purge ON at 15 C/min to 265 C and held at 265 C for 5 min. Voltage scan MS data were recorded at resolution, 5% valley definition. (b) Quadrupole GC/MS instrument. Finnigan (San Jose, CA) 4023T quadrupole mass spectrometer coupled to a Hewlett Packard 5890 Series II gas chromatograph equipped with dual injection ports and electronic pressure control. Extracts in aqueous acidic or ethyl acetate solution were analyzed as received with the quadrupole GC/MS instrument. Helium carrier gas flow was set at 58 cm/s (2.66 ml/min) with a 55 C oven temperature, vacuum compensation and constant flow controls turned ON, and split flow set at 30 ml/min. Injections of apple juice extract in ethyl acetate used to prepare Figure 1 were splitless via the split/splitless injector; the capillary column and transfer line temperature settings were the same as those used with the high-resolution GC/MS instrument. The ion source temperature of the quadrupole mass spectrometer was 200 C. The quadrupole GC/MS instrument was modified (10) to accept capillary columns up to 0.53 mm id. A 15 m 0.53 mm id capillary column with a 1.5 µm film of bonded phase methyl polysiloxane was used for on-column injection of the aqueous acidic extract via the programmable injection port to obtain the data shown in Figure 2. Helium carrier gas flow was set at 25 cm/s with vacuum compensation and constant flow OFF to produce an indicated 213 cm/s (17.9 ml/min) with a 60 C oven temperature and vacuum compensation and constant flow ON. Temperatures for the on-column injections were 63 C injector, oven track ON, and 260 C transfer lines. The GC oven was held at 60 C for 1 min after on-column injection, heated at 15 C/min to 260 C, and held at 260 C for 5 min. Figure 1. (A) Reconstructed ion profiles extracted from full-scan data. (B) Multiple ion detection data recorded for the same test portion. Improved signal-to-noise ratios with multiple ion detection are evident in the area counts recorded for the ion profiles in A and B: m/z 108, (A) 45 and (B) 1076; m/z 110, (A) 59 and (B) 943; m/z 124, (A) 30 and (B) 653; m/z 136, (A) 1546 and (B) ; m/z 154, (A) 199 and (B) 4012; total ion, (A) 7808 and (B) Quadrupole data were acquired in the negative ion mode under INCOS data system control. LC patulin quantitation was 68 g/l (68 ppb).

3 106 ROACH ET AL.: JOURNAL OF AOAC INTERNATIONAL VOL. 83, NO. 1, 2000 Patulin eluted between 6 and 10 min with the GC conditions and columns described in this paper. Data acquisition with the quadrupole instrument was started after the GC run was under way; therefore, the indicated retention times for patulin were shifted by 3 5 min. (c) Ion trap GC/MS instrument. Thermoquest (San Jose, CA) GCQ ion trap GC/MS system with version 2.2 software supporting tuning and calibration in the negative ion CI mode. A30m 0.25 mm id capillary column with a 0.25 µm film of bonded phase trifluoropropylmethyl polysiloxane was used for splitless injections of ethyl acetate solutions in the split/splitless injector. All injections were made with an autoinjector under data system control. Constant helium carrier flow was set at 40 cm/s (1.21 ml/min). Temperatures were 260 C injector, transfer lines 260 C, and ion source 200 C. Column temperature was programmed from 60 to 260 C 1 min after injection at a rate of 20 C/min. (d) Syringes. Hamilton (Reno, NV) 701-SN 10 µl, 7.5 cm needle for use with 4 mm glass, double-gooseneck, splitless injector insert used in all 3 GC/MS systems. Hewlett Packard µl, 5 cm 23-/26 ga tapered needle, for on-column injection into 0.53 mm id capillary GC column. (e) Glass tubes mm disposable glass culture tubes with Teflon-lined screw caps. Figure 2. Confirmatory data for patulin in aqueous acidic extract of apple juice concentrate. The inset ion profiles from top to bottom are total ion, m/z 154, and m/z 136. Patulin data were obtained with 0.8 s repetitive scans of the quadrupole from 60 to 160 daltons in the negative ion mode. Filament emission was 0.25 ma at 70 ev. Data acquisition with the Teknivent data system began 3 min after injection; therefore, the true retention time of patulin was 6.59 min. The mass spectrum of the analyte at 6.59 min matches that of standard patulin. LC patulin quantitation was 2867 g/l (2.9 ppm).

4 ROACH ET AL.: JOURNAL OF AOAC INTERNATIONAL VOL. 83, NO. 1, Figure 3. Ion trap confirmatory negative ion CI data for patulin in ethyl acetate extract of apple juice, obtained with a 5% phenyl/methyl polysiloxane capillary GC column and split injection with 30:1 split ratio. Ion profiles from top to bottom are m/z 154, m/z 136, m/z 108, and total ion (TOT). The inset mass spectrum of the analyte recorded at 9.46 min matches that of standard patulin. LC quantitation indicated 16 ppm patulin. (f) Pasteur pipet. 9 in. length (ca 23 cm). (g) Kuderna Danish concentrators. 250 ml, with Vigoro distilling column and 1 ml graduated receiving flasks. (h) Vortex mixer. Equivalent tube shaker is also suitable. Reagents and Solutions (a) Methane. Matheson ultra high purity (East Rutherford, NJ). (b) Ethyl acetate. LC grade; UV cutoff, 252 nm. (c) Acetonitrile. LC grade. (d) Sodium carbonate solution. Dissolve 1.5 g sodium carbonate (reagent grade) in 100 ml deionized water. (e) Acetic acid solution. Water adjusted to ph 4 with acetic acid (reagent grade). (f) Patulin standard spiking solution (80 ppm). Weigh 8 mg pure crystalline patulin into 100 ml volumetric flask. Dissolve patulin in acetonitrile. Dilute to volume with acetonitrile. (g) Patulin standard stock and working solutions in ethyl acetate. Weigh 2 mg pure crystalline patulin into 10 ml volumetric flask. Dissolve patulin and dilute to volume with ethyl acetate to make 200 ppm stock solution. Dilute 1 ml 200 ppm stock solution to 50 ml to make 4 ppm (4000 ppb) patulin solution. Dilute 1 ml 4 ppm solution to 10 ml to make 400 ppb solution. Dilute 1 ml 400 ppb solution to 5 ml to make 80 ppb solution. Dilute 1 ml 400 ppb solution to 10 ml to make 40 ppb solution. Dilute 1 ml 40 ppb solution to 5 ml to make 8 ppb solution. Dilute 1 ml 40 ppb solution to 10 ml to make 4 ppb solution. Dilute 1 ml 4 ppb solution to 10 ml to make 0.4 ppb solution. Use ethyl acetate solvent as 0 ppb solution. MS Determination The magnetic sector instrument was used to obtain high-resolution electron ionization (EI) data for patulin. The quadrupole mass spectrometer provided EI and CI data with positive ion and negative ion modes of detection. Methane

5 108 ROACH ET AL.: JOURNAL OF AOAC INTERNATIONAL VOL. 83, NO. 1, 2000 pressure in the quadrupole ion source in the CI mode produced a manifold pressure of torr. The ion trap was operated in EI, positive ion CI, and negative ion CI modes, according to its operating manual, in analyses for patulin. Negative ion CI quadrupole and ion trap data were acquired over a mass range of daltons for analyses of juice extracts containing incurred and spiked residues of patulin. Confirmation of patulin identity was based on comparison of analyte signals with those of patulin and a reagent blank analyzed under identical GC/MS conditions. Results and Discussion Several researchers have reported the confirmation of patulin in apple juice by GC/MS following formation of a derivative of patulin (8, 9, 11 13). The derivatization step may result in sample loss through evaporation, incomplete reaction, or decomposition before analysis. The reagents typically react to form products detected as additional peaks in the chromatogram. Tarter and Scott (9) found it necessary to identify these extraneous peaks as well as the patulin derivative in their data. Components in the sample other than patulin may also react with the derivatizing reagent. The result is a series of responses in derivatized extracts that have no counterpart in an untreated extract (11, 12). Mortimer et al. reported that silylation precluded confirmation by GC/MS in the EI mode because of significant silylated interferences (12). The mass spectra of patulin derivatives provide additional ions of mass higher than that of the molecular weight of patulin itself. For example, the molecular weight of patulin trimethylsilyl ether is 226, and the molecular weight of patulin heptafluorobutyrate is 350. The higher molecular weights of the derivatives provide the possibility of the occurrence of ions in the derivative spectra that are further away from low mass chemical noise than are all of the ions in the mass spectrum of patulin. Unfortunately, many of the ions in the spectra of patulin derivatives arise from the derivative rather than the patulin portion of the molecule. As such, they do little more than confirm the derivatized molecular weight because they are not diagnostic for patulin. Patulin is a lactone with a nominal molecular weight of 154. The EI mass spectrum of patulin contains an abundant molecular ion at m/z 154 and ions resulting from losses of the elements of water (m/z 136) and carbon monoxide (m/z 126). The base peak at m/z 110 has an elemental composition of C 5 H 2 O 3, suggesting loss of C 2 H 4 O from the pyran ring as a neutral species or stepwise losses of CH 2 O and CH 2 from the molecule. The elemental composition of C 5 H 5 O 2 for m/z 97 is consistent with loss of the elements of C 2 HO 2 from the lactone ring of the molecule. Electrons with an average energy of 70 ev ionize the analyte in the EI mode of analysis. Excess energy imparted to the molecule during ionization is available for fragmentation of the molecular ion. In contrast, CI occurs with the making and breaking of chemical bonds or the capture of low-energy electrons. The energy scale for CI processes is significantly lower than what is available under EI conditions. The reactive plasma in a CI source chemically reacts with the analyte and also acts to reduce the energy of ions and electrons within the source through multiple collisions with neutral reagent gas molecules (10). Positive ion CI processes occur on an energy scale of 10 ev and electron capture negative ion CI proceeds at 2.5 ev. As a result, positive ion CI spectra show less fragmentation than do EI spectra, and fragmentations after electron capture under negative ion CI conditions are limited to pathways that require very little energy (14). The reduced fragmentation can increase the sensitivity of the assay because the total signal for the analyte is contained in a small set of ions provided by the most favorable fragmentation pathways. However, the improved sensitivity may be of little use if the set of ions is not diagnostic for the analyte. The methane positive ion CI spectrum of patulin obtained with the quadrupole instrument consists primarily of an [M+H] + ion at m/z 155 as the base peak and a less abundant ion at m/z 137. The methane negative ion CI spectrum of patulin contains an abundant molecular anion, a base peak corresponding to the loss of water at m/z 136, and ions of < 30% relative abundance at m/z 64, 108, 110, 124, and 126. Similar negative ion CI spectra are obtained with the ion trap. In 1993, a survey of fresh, fresh processed, concentrate, frozen concentrate, and fresh processed plus concentrate apple juice products for patulin required MS confirmation of presumptive findings provided by LC with UV detection. Underivatized patulin was found to be amenable to capillary GC. Full-scan negative ion CI was found to be more sensitive for patulin, and the data contained fewer extraneous co-extractive signals than did full-scan EI data; therefore, patulin levels in excess of 500 µg/l (ppb) were identified by full-scan negative ion CI GC/MS, but the full-scan protocol faltered for extracts in which the patulin level was 50 µg/l (ppb). Multiple ion detection of the negative ion patulin signal was required. For example, Figure 1A shows reconstructed ion profiles extracted from full-scan negative ion data for a juice extract containing incurred patulin at 68 ppb, as measured by LC with UV detection. Figure 1B shows multiple ion data recorded for the same extract. The quadrupole instrument used to develop the negative ion assay for patulin was recently replaced with an ion trap. Multiple ion detection and 0.53 mm id columns were not appropriate analytical options with a trap. Ion trap analyses for patulin were performed with mass scans of 60 to 160 daltons for EI and negative ion CI, mass scans of 65 to 205 daltons for positive ion CI, and capillary columns of 0.32 or 0.25 mm id. Use of the trap in analyses for patulin revealed a problem with the chromatography of patulin on low-bleed 5% phenyl/methyl polysiloxane columns. Solving this chromatographic problem significantly enhanced the sensitivity of the assay for patulin. Ion traps are very sensitive instruments, but they function just like the buckets in a water wheel. After ionization, the trap fills with ions. The ions in the trap are mass analyzed and scanned out to the detector. The trap then undergoes more cycles of ion collection, mass analysis, and detection to complete

6 ROACH ET AL.: JOURNAL OF AOAC INTERNATIONAL VOL. 83, NO. 1, Figure 4A. EI ion trap data recorded for 1 ng patulin. Figure 4B. Negative ion CI trap data recorded for 1 ng patulin.

7 110 ROACH ET AL.: JOURNAL OF AOAC INTERNATIONAL VOL. 83, NO. 1, 2000 Figure 4C. Positive ion CI trap data recorded for 10 ng patulin. The qualitative identification limits with a 5% phenyl/methyl polysiloxane capillary GC column were 10 ng for positive ion CI, 4 10 ng for EI, and 1 ng for negative ion CI. the analysis. A trap can hold only a finite number of ions in each scan cycle. If most of the ions sampled by the trap arise from column bleed or co-extractive materials rather than analyte, the analyte signal will be suppressed by these competing ion species. Low-bleed GC columns minimize column-bleed contributions to chemical noise in the ion trap, but low-bleed columns may retain patulin more strongly and cause the patulin peak to tail (Figures 3, 4). Increasing the heating rate to 20 C/min narrowed the patulin peak width, but the real solution was to find a stationary phase that was appropriate for patulin. Bonded phase trifluoropropylmethyl polysiloxane provides satisfactory chromatography of underivatized patulin in ethyl acetate solutions, as evidenced by a marked increase in the sensitivity of the assay for patulin. Sensitivity is enhanced because the detector receives more molecules of patulin per unit of time when the analyte is confined in a narrow GC peak rather than dispersed in a broad GC peak. Comparisons of EI and CI ion trap sensitivity with a 5% phenyl/methyl polysiloxane GC column agreed with the earlier quadrupole data, in that negative ion CI provided the best relative limit of detection (Figure 4A C). However, changing to the trifluoropropylmethyl polysiloxane capillary GC column redefined these limits. A 25 ng injection of patulin overloaded the column. A 10 ng injection overloaded the detector. A calibration curve of ion intensity recorded for m/z 136 from 400 fg to 400 pg was linear. However, the qualitative identification limit was set at 8 pg because data recorded for 8 pg could be used to reliably identify patulin, whereas the 4 pg data were marginal. Data points below 8 pg were discarded, and the curve remained linear (R = , n = 10). The standard deviation (SD) for 4 injections of 400 pg spaced over 8 h was 3.8%, and the SD for 4 injections of 40 pg spaced over 8 h was 12.4%. Portions of apple juice known to contain <5 ppb patulin were spiked with 40 ppb patulin by adding 25 µl ofthe 80 ppm solution of patulin in acetonitrile to each 5 ml juice portion to make 40 ppb patulin-spiked test portions. The 5 ml portions were extracted twice with 10 ml portions of ethyl acetate by mixing the spiked juice and ethyl acetate in mm disposable glass culture tubes with a Vortex mixer for 1 min. Cleanup of the ethyl acetate extract with sodium carbonate solution was omitted to intentionally create less than optimally clean extracts for analysis. The organic layer was drawn off with a Pasteur pipet and combined before drying the ethyl acetate extract with 1 g anhydrous granular sodium sulfate. A 10 ml portion of ethyl acetate extract was withdrawn and concentrated to 1 ml with a Kuderna Danish apparatus on a steam bath. Representative ion trap data for the GC/MS analyses of the spiked extracts are shown in Figure 5.

8 ROACH ET AL.: JOURNAL OF AOAC INTERNATIONAL VOL. 83, NO. 1, Figure 5. Negative ion CI trap data obtained with a trifluoropropylmethyl polysiloxane capillary CG column for ethyl acetate extract of apple juice. The negative ion CI patulin qualitative identification limit with this capillary column was 8 pg. Sodium carbonate extract cleanup was not used. Negative ion CI of patulin proceeds through electron capture. Because not all molecules undergo electron capture, the negative ion process serves to increase the selectivity of the assay by selectively ionizing patulin. A mass scan assay further enhances selectivity for patulin by providing a record for all ions in the mass range rather than a record for a small set of preselected ions found in the spectrum of patulin. In this case, the absence of co-extractive signals in the data at the retention time of patulin are as important as the patulin signal itself in proving that the response is due to patulin. The intended application of the technique was the rapid and simple confirmation of presumed patulin at levels at which consumption of the product might result in immediate harm rather than chronic exposure of the consumer. The better sensitivity of newer MS instrumentation, such as the ion trap in combination with a better GC column, makes it possible to apply the technique to significantly lower patulin levels. However, quantitation is still based on LC rather than MS. A satisfactory quantitative MS assay requires a suitable internal standard such as stable isotope-labeled patulin (15). This requirement is particularly true for the ion trap, because the observed analyte signal is influenced by the population of ions in the trap during the measurement. The signal recorded for a trap full of patulin ions is larger than the patulin signal recorded for the same amount of patulin diluted with ions from column bleed and co-extractives. A labeled internal standard compensates for this variation. Quantitation based on an external calibration standard or process standards such as nitrobenzene (13) or hexachlorobenzene (7) will not detect or correct this problem. The LC quantitation method has been successfully studied collaboratively. Values derived by LC should be accepted as valid on the basis of the collaborative results until commercially available labeled internal standards permit an independent determination of patulin by GC/MS. Acknowledgments We thank Denis Andrzewjewski and Martin J. Stutsman (U.S. Food and Drug Administration, Washington, DC) for their useful comments in the preparation of this manuscript, and David A. Nortrup (FDA) for the use of his sample preparation area and glassware.

9 112 ROACH ET AL.: JOURNAL OF AOAC INTERNATIONAL VOL. 83, NO. 1, 2000 References (1) Codex Committee on Food Additives and Contaminants, 29th Session (March 17 21, 1997) The Hague, The Netherlands, Position Paper on Patulin, Agenda item 16(d), CX/FAC 97/23-Part II, January 1997 (2) Van Egmond, H.P., & Dekker, W.H. (1995) Nat. Toxins 3, (3) Sharma, R.P. (1993) J. Dairy Sci. 76, (4) Rodricks, J.V., Hesseltine, C.W., & Mehlman, M.A. (1977) Mycotoxins in Human and Animal Health, Pathotox Publishers, Park Forest South, IL, pp (5) Wood, G.E., & Trucksess, M.W. (1999) Mycotoxins in Agriculture and Food Safety, K.K. Sinha & D. Bhatnagar (Eds), Marcel Dekker, New York, NY, pp (6) Brause, A.R., Trucksess, M.W., Thomas, F.S., & Page, S.W. (1996) J. AOAC Int. 79, (7) Llovera, M., Viladrich, R., Torres, M., & Canela, R. (1999) J. Food Prot. 62, (8) Rosen, J.D., & Pareles, S.R. (1974) J. Agric. Food Chem. 22, (9) Tarter, E.J., & Scott, P.M. (1991) J. Chromatogr. 513, (10) Roach, J.A.G. (1998) in Spectral Methods in Food Analysis: Instrumentation and Analysis, M.M. Mossoba (Ed.), Marcel Dekker, New York, NY, pp (11) Kellert, M., Baltes, W., Blaas, W., & Wittkwoski, M., (1983) Fresenius Z. Anal. Chem. 315, (12) Mortimer, D.N., Parker, I., Shepherd, M.J., & Gilbert, J. (1985) Food Addit. Contam. 538, (13) Sheu, F., & Shyu, Y.T. (1999) J. Agric. Food Chem. 47, (14) Brumley, W.C., & Sphon, J.A. (1987) in Applications of Mass Spectrometry in Food Science, J. Gilbert (Ed.), Elsevier Applied Science, New York, NY, pp (15) Rychlik, M., & Schieberle, P. (1998) J. Agric. Food Chem. 46,