Solid-phase extraction followed by dispersive liquid–liquid microextraction for the sensitive determination of selected fungicides in wine

Solid-phase extraction followed by dispersive liquid-liquid microextraction for the sensitive determination of selected fungicides in wine R. Montes, I. Rodríguez, M. Ramil, E. Rubí, R. Cela Departamento de Química Analítica, Nutrición y Bromatología, Instituto de Investigación y Análisis Alimentario, Universidad de Santiago de Compostela, Santiago de Compostela 15782, Spain. Abstract A novel approach for the determination of seven fungicides (metalaxyl-M, penconazole, folpet, diniconazole, propiconazole, difenoconazole and azoxystrobin) in wine samples is presented. Analytes were extracted from the matrix and transferred to a small volume of a high density, water insoluble solvent using solid-phase extraction (SPE) followed by dispersive liquid-liquid microextraction (DLLME). Variables affecting the performance of both steps were thoroughly investigated (metalaxyl-M was not included in some optimisation studies) and their effects on the selectivity and efficiency of the whole sample preparation process are discussed. Under optimised conditions, 20 mL of wine were first concentrated using a reversed-phase sorbent and then target compounds were eluted with 1 mL of acetone. This extract was mixed with 0.1 mL of 1,1,1trichloroethane (CH3CCl3) and the blend added to 10 mL of ultrapure water. After centrifugation, an aliquot (1-2 µL) of the settled organic phase was analyzed by gas chromatography (GC) with electron capture (ECD) and mass spectrometry (MS) detection. The method provided enrichment factors (EFs) around 200 times and an improved selectivity in comparison to use of SPE as single sample preparation technique. Moreover, the yield of the global process was similar for red and white wine samples and the achieved limits of quantification (LOQs) (from 30 to 120 ng L-1 and from 40 to 250 ng L-1, for GC-ECD and GC-MS, respectively) were low enough for the determination of target species in commercial wines. Among compounds considered in this work, metalaxyl-M and azoxystrobin were found in several wines at concentrations from 0.8 to 32 ng mL-1. Keywords: Dispersive liquid-liquid microextraction; solid-phase extraction; fungicides; wine; sample preparation; gas chromatography. This is the postprint(accepted manuscript) version of the article published in Journal of Chromatography A https://doi.org/10.1016/j.chroma.2009.05.048 This manuscript version is made available under de CC-BY-NC-ND 4.0 license hppt://creativecommons.org/licenses/by-nc-cd/4.0 1. Introduction Fungicides represent one the most relevant groups of agrochemicals applied to vineyards. These compounds are sprayed directly on grapes and leaves to prevent and to control the attack of fungi which reduce the productivity of vines as well as the quality of grapes and wine [1]. Depending on several factors, such as dosage, interval between field application and harvest, operations involved in the wine-making process and particularly the physico-chemical properties of fungicides, residues of these compounds and/or their by-products might reach commercial wines [2-4]. Although, with a few exceptions [5], the presence of fungicides in wine is not legislated, it is an issue of concern for consumers and producers. Consequently, the development of improved analytical methodologies to measure their residues in wine, as well as to evaluate their fate through the wine-making process, is a topic of increasing interest and a field of active research. Chromatographic techniques are the most resorted for the determination of agrochemical residues in wine. The complexity of this matrix and the need to achieve limits of quantification (LOQs) in the low ng per mL range makes necessary to optimize a previous sample preparation step. Solid-phase extraction (SPE) [6,7] has been proposed for the concentration of wine as alternative to liquid-liquid extraction (LLE) based methodologies [8]. In most applications, a volume of sample in the range from 10 to 50 mL is passed through a reversed-phase SPE sorbent and then analytes are recovered using an organic solvent compatible with the further determination step. SPE methodologies often render high extraction yields; however, their selectivity is relatively low, unless an additional clean-up, based on the use of normal-phase materials, is carried out [9]. Solidphase microextraction (SPME) has been also applied to the determination of fungicides, belonging to different chemical classes, in wine using gas chromatography (GC) [10-12], liquid chromatography (LC) [13] and even capillary electrophoresis [14,15]. Although SPME normally provides a higher selectivity than SPE, the ethanol content of wine reduces significantly its extraction efficiency when compared to water samples [16,17]; moreover, the kinetics of the extraction is relatively slow and differences in the yield of the process, depending on the wine matrix, have been reported [18]. Consequently, quantification relies on the time consuming standard addition method. Above limitations are also common to stir-bar sorptive extraction (SBSE), whose applicability is restricted to low polar fungicides, showing a high affinity for the polydimethylsiloxane sorbent [19]. Conversely to SPME, the suitability of liquid-phase microextraction (LPME) for the extraction of fungicides from wine samples remains mostly unexplored. The performance of LPME is expected to be controlled by the chosen configuration [20]. Theoretically, membrane based modalities provide moderate to high selectivities, at the expense of very slow extraction kinetics, particularly when non-porous membranes are used [19]. On the other hand, the stability of the hanged extractant phase could be comprised in the single-drop microextraction (SDME) format [21]. Some of these drawbacks have been overcome by the dispersive liquid-liquid microextraction (DLLME) technique, first reported by Assadi and co-workers [22]. This modality of LPME uses a binary mixture of a high density solvent (extractant) and a water soluble one (disperser) to isolate and concentrate trace amounts of organic compounds from aqueous matrices. In most cases, DLLME has been applied directly to water samples [23-26]; moreover, it can be also combined with SPE [27] and LLE [28] to improve the selectivity of the sample preparation process and/or to reduce the achieved LOQs for complex matrices. This work evaluates the suitability of DLLME, in combination with reversed-phase SPE, for the extraction of a selected group of fungicides, belonging to four different chemical classes (phenylamides, phthalimides, azoles and strobilurines), from red and white wine samples. The effects of different parameters on the efficiency and selectivity of the global sample preparation method are thoroughly discussed, except in case of metalaxyl-M, which was not included in some steps of the optimization process. Figures of merit for the proposed approach are reported using GC with electron capture (ECD) and mass spectrometry (MS) detection; moreover, data regarding the levels of target compounds in some commercial wines are also given. 2. Experimental 2.1. Solvents, standards and material HPLC-grade methanol was purchased from Merck (Darmstadt, Germany), trace analysis quality acetone, carbon disulphide (CS2), carbon tetrachloride (CCl4), chlorobenzene (C6H5Cl), chloroform (CHCl3) and 1,1,1-trichloroethane (CH3CCl3) were obtained from Merck and Aldrich (Milwaukee, WI, USA). Sodium chloride was provided by Aldrich. Standards of metalaxyl-M (99.0%), penconazole (99.1%), folpet (99.9%), diniconazole (99.8%), propiconazole (98.6%), difenoconazole (97.0%) and azoxystrobin (99.9%) were acquired from Aldrich and Merck. Their chemical structures, CAS numbers and some relevant properties are summarized in Table 1. Individual solutions of each fungicide were prepared in methanol. Further dilutions and mixtures of them, used to fortify wine samples as well as wine extracts, were made in acetone. Calibration standards, employed to assess the efficiency of the DLLME process, were dissolved in CH3CCl3. OASIS HLB (60 mg) SPE cartridges were acquired from Waters (Milford, MA, USA). Glass fibre pre-filters were purchased from Millipore (Bedford, MA, USA). Glass tubes (12 mL volume) with conic bottom and screw caps were provided by Afora (Barcelona, Spain). Number Compound Structure CAS number pKa log Kow 1 Metalaxyl-M 70630-17-0 1.41 2.2 O O O O N R O O O O N R 2 Penconazole 66246-88-6 2.83 3.7 3 Folpet 133-07-3 -3.34 2.9 4 Diniconazole 83657-24-3 2.19 4.2 5 Propiconazole 60207-90-1 3.06 3.9 6 Difenoconazole 119446-68-3 3.06 4.9 7 Azoxystrobin 131860-33-8 -0.67 5.1 Table 1. Chemical structure, CAS number, pKa and log Kow values of target species. 2.2. Samples and sample preparation Spiked and non-spiked wine samples, acquired from local supermarkets, were used in this work. Sample preparation conditions were optimized with aliquots of a pooled matrix prepared with different red wines ( Tempranillo, Grenache and Cabernet Sauvignon ). After filtration, they were diluted with ultrapure water (1:1) and passed through a SPE cartridge, previously conditioned with the elution solvent and ultrapure water (5 mL each). Then, cartridges were rinsed with 5 mL of water, dried for 15 min using a gentle stream of nitrogen and eluted with a suitable solvent. Breakthrough studies were performed passing the spiked wine samples through two cartridges connected in series. After the enrichment step they were disconnected and processed independently. Cl NCl N N Cl NCl N N O O Cl Cl Cl N S O O Cl Cl Cl N S Cl Cl N N OH N Cl Cl N N OH N Cl Cl N N O O N Cl Cl N N O O N Cl OCl N N O O N Cl OCl N N O O N N N O N O O O ON N O N O O O O Optimization of DLLME parameters was carried out with a pool of SPE extracts, from red wine, fortified after the SPE step. A fraction of this matrix was mixed with the extractant (carbon disulphide or a high density chlorinated solvent) and transferred to the barrel of a 2 mL disposable polypropylene syringe, furnished with a stainless steel needle. Then, this blend was added rapidly to ultrapure water contained into a conical bottom glass tube. The ternary mixture was shaken and centrifuged to allow phases separation and the volume of the settled drop measured with a 100 µL micro-syringe. Efficiency of the DLLME step was defined as the ratio between the mass of each species in the settled phase and that added to the pooled SPE extract from un-spiked wine samples. The first was calculated as the product between the volume of the sedimented phase and the concentration of fungicides, which was established by external calibration against standards prepared in the same solvent used as extractant in the DLLME process. Overall enrichment factors (EFs) were defined as the ratio between sample and final extract volumes multiplied by the efficiencies (recoveries) of SPE and DLLME steps. Under final conditions, the sample intake was limited to 20 mL of filtered wine. Analytes were recovered from the OASIS HLB sorbent with 1 mL of acetone, which was mixed with 0.1 mL of CH3CCl3. This blend was injected into 10 mL of water to obtain a cloudy mixture and centrifuged at 3600 rpm (1100 g) for 3 min. Around 50 µL of the sedimented phase (total volume 78 µL) were transferred to a 0.2 mL insert and 1-2 µL injected in the chromatographic system using an autosampler. 2.3. Instrumentation An Agilent (Wilmington, DE, USA) 6890 model gas chromatograph (GC) equipped with a microelectron capture detector (micro-ECD) was used during optimisation of most of the sample preparation conditions. As metalaxyl is hardly detected by ECD, the influence of some parameters on the extraction of this specie were not evaluated. The GC-ECD system was furnished with autosampler, split/splitless injector and electronic pressure control. Analytes were separated with an Agilent HP-5 type capillary column (30 m x 0.32 mm i.d., df: 0.25 µm) operated at a constant helium flow of 1.4 mL min-1. The column temperature program was as follows: 1 min at 100 ºC, ramp at 20 ºC min-1 to 295 ºC (held for 10 min). Injector and detector were set at 280 and 300 ºC, respectively. Nitrogen was used as make-up gas in the micro-ECD at a constant flow of 40 mL min1. Injections (1-2 µL) were made in the pulsed splitless mode, considering an initial pressure of 35 psi for 1.2 min, with the solenoid valve passing to the split position after 1 min. Performance of the optimised method was assessed by GC-ECD and GC-MS. The latter system was comprised of a Varian (Walnut Creek, CA, USA) 450 GC instrument connected to an ion-trap Varian 240 mass spectrometer (MS). Separations were carried out in a Varian Factor Four column (30 m x 0.25 mm i.d., df: 0.25 µm) operated at a constant helium flow of 1.2 mL min-1. The temperature of the inlet and the injection mode were the same as in the GC-ECD system. The oven temperature was programmed as follows: 1 min at 80ºC, first rate at 12 ºC min-1 to 245 ºC and then at 25 ºC min-1 to 285 ºC (held for 10 min). Transfer line, electron impact ionization source and trap were set at 285, 200 and 150 ºC, respectively. MS spectra were recorded in the range from 80 to 400 m/z units, considering the most intense ions of each compound for quantitative purposes. The NIST MS pesticide database was used to identify the presence of additional fungicides in unspiked commercial wines. 3. Results and discussion 3.1. Sample concentration strategy In the early steps of this research, the suitability of DLLME and SPE as isolated techniques, for the extraction of fungicides from wine was assessed. As regards DLLME, assays were performed using two different volumes of wine (5 and 10 mL) and several combinations of extractants and dispersers. Although Fariña and co-workers [29] have reported the direct application of this technique to the extraction of phenolic species from red wine, whatever the chosen combination of disperser and extractant, as well as the pH of the sample, some components of the matrix precipitated at the bottom of the extraction tube, together with the extractant, preventing the injection of settled phase in the GC system. The above referred problem was particularly relevant for red wines; whereas, for white ones, it could be alleviated by using a large volume of extractant. For this latter matrix, the best separation between the upper wine layer and the settled organic phase, as well as the lowest amount of precipitate in the latter, was achieved using acetone and chloroform as disperser and extractant, respectively. On the other hand, SPE was also evaluated as sample concentration technique. Wine was previously diluted (1:1) with ultrapure water and then concentrated using a reversed-phase OASIS HLB cartridge. The complexity of the resulting extracts was controlled by the type of wine and the elution solvent. Using acetone or methanol, the extracts from red wines presented a high content of pigments, whereas pale, yellowish ones were obtained with ethyl acetate. In spite of this, relatively complex GC-ECD chromatograms were attained, reducing the signal to noise (S/N) ratio of peaks corresponding to target species, Fig. 1A. min 4 6 8 10 12 14 16 Hz 0 100000 200000 300000 400000 2 3 4 5 6 7 A B Fig. 1. GC-ECD chromatograms for 50 mL of a pooled red wine sample spiked at 5 ng mL-1. A, SPE using ethyl acetate as elution solvent with concentration of the extract to 0.2 mL. B, SPE followed by DLLME, acetone was used as elution solvent and CH3CCl3 (0.1 mL) as extractant. Same peak numbers as in Table 1. Finally, the tandem use of SPE and DLLME avoided some of the above drawbacks. In this case, SPE cartridges were eluted using a water miscible solvent (e.g. acetone), this extract was mixed with a small volume of a high density solvent and the mixture added to 10 mL of water. This combination provided transparent extracts, even for red wine samples, without the presence of any precipitate in the sedimented extract. In addition, GC-ECD chromatograms presented a lower baseline noise (Fig. 1B), confirming the improvement in the selectivity of the sample preparation process. 3.2. Optimization of sample preparation conditions Optimization of extraction conditions was carried out with a pooled sample of red wines, evaluating first the parameters which controlled the efficiency of the SPE step and then, those related to the DLLME process. In the first case, SPE extracts (1 mL) were mixed with 0.1 mL of CH3CCl3, as extractant, and added to 10 mL of water in order to obtain an emulsion which promoted the transference of the analytes to the dispersed drops of chlorinated solvent. 3.2.1. SPE parameters Breakthrough studies demonstrated the capacity of the OASIS HLB sorbent to concentrate up to 50 mL of wine, previously diluted with ultrapure water (1:1) to avoid the negative effect of ethanol on the yield of the retention [7], without noticeable losses of any species. In the elution step, consecutive fractions (1 mL each) of methanol or acetone were collected. In both cases, analytes were detected just in the first 1 mL fraction. Although reddish extracts were obtained with both solvents; visually, the lower level of co-extracted pigments corresponded to acetone. When the SPE extract was combined with 0.1 mL of CH3CCl3, in the further DLLME process, a slightly larger sedimented phase were observed for acetone than for methanol (78 vs. 70 µL, respectively). In addition, for methanol a whitish interface was noticed between the settled drop of CH3CCl3 and the upper aqueous phase in the DLLME tube; whereas, a better phase separation was noticed for acetone extracts. On the basis of these comments, acetone was chosen to elute target compounds from the OASIS HLB cartridge. Table 2 summarizes the recoveries of the SPE process for red wine. Depicted data were obtained comparing the responses obtained for spiked wine samples with those measured for extracts, corresponding to un-spiked aliquots of the same wine, fortified after the SPE step. Although considered compounds present significant differences in their polarities (log Kow values from 2.2 to 5.1 units) and some of them, e.g. propiconazole and difenoconazole, show pKa values close to the pH of wine (3.1-3.5 units), acceptable recoveries (over 82%) were attained for all of them in the SPE process. Compound Recovery (%) ± standard deviation Metalaxyl-M 82.3 ± 3.9 Penconazole 91.8 ± 1.0 Folpet 93.7 ± 1.9 Diniconazole 86.7 ± 1.1 Propiconazole 90.5 ± 3.1 Difenoconazole 97.6 ± 5.6 Azoxystrobin 84.1 ± 4.7 Table 2. Absolute recoveries of the SPE process for 50 mL aliquots of a pooled red wine sample. Addition level 10 ng mL -1 , n= 3 replicates 3.2.2. DLLME parameters 3.2.2.1. Type and volume of extractant Performance of DLLME is mainly determined by the type and volume of extractant. In this work, CHCl3, CCl4, CH3CCl3, C6H5Cl and CS2 were evaluated as potential extractants. Aliquots (0.1 mL) of these solvents were mixed with 1 mL fractions of a pooled SPE extract (corresponding to red wine) in acetone, and added to 10 mL of ultrapure water using a polypropylene syringe. CHCl3 and CCl4 led to settled drops with volumes of 20 and 35 µL, respectively; whereas, values between 78 and 80 µL were measured for CH3CCl3, C6H5Cl and CS2. Considering that larger drops are easier to handle, the three latter solvents were selected for further experiments. Fig. 2 compares the results obtained with these extractants for triplicate assays, using GC-ECD as detection technique. Depicted responses were normalized to those attained for CH3CCl3. In the case of propiconazole and difenoconazole the sum of peak areas for both isomers was used as variable response. For all compounds, the lowest responses corresponded to C6H5Cl. Comparison between CS2 and CH3CCl3 showed that the chlorinated solvent provided a better repeatability (Fig. 2), as well as a lower baseline noise in the corresponding chromatograms, figure not shown. Thus, CH3CCl3 was maintained as extractant. 0% 20% 40% 60% 80% 100% 120% Penconazole Folpet Diniconazole Propiconazole Difenoconazole Azoxystrobin CS 2 CH 3 CCl 3 C 6 H 5 Cl Normalized response Fig. 2. Effect of the extractant on the efficiency of the DLLME process. Normalized values to those achieved using CH3CCl3, n=3 replicates. Extractant and settled phase volumes were 100 and 78-80 µL, respectively. Table 3 shows the efficiency (absolute recoveries) of the DLLME step using two different volumes of CH3CCl3, 60 and 100 µL. A 10-14% increase in the yield of the extraction, depending on the fungicide, was noticed when the volume of extractant rose from 60 to 100 µL. On the other hand, the size of the settled phase passed from 40 to 78 µL; therefore, nearly twice EFs were achieved with the lower volume of CH3CCl3. In spite of this, it was decided to fix the volume of CH3CCl3 at 0.1 mL to facilitate storage and re-injection of sample extracts. Compound Efficiency (%) ± SD 60 µLa (40 µL) b 100 µLa (78 µL) b Metalaxyl-M 60.2 ± 4.3 74.0 ± 3.2 Penconazole 66.8 ± 3.5 80.7 ± 3.5 Folpet 75.1 ± 1.7 83.5 ± 1.0 Diniconazole 72.9 ± 4.3 84.9 ± 1.5 Propiconazole 73.4 ± 2.8 83.5 ± 1.9 Difenoconazole 91.3 ± 2.0 101.9 ± 3.2 Azoxystrobin 89.7 ± 3.2 98.1 ± 6.5 Table 3. Efficiency of the DLLME step using two different volumes of CH3CCl3, n=3 replicates. a Extractant volume; b Settled phase volume 3.2.2.2. Salt addition and water pH The effect of the ionic strength on the yield of the DLLME step was investigated considering 10 mL of water with 3 different concentrations of NaCl, and using GC-ECD as detection technique. As improved by using both techniques in tandem, overcoming some potential limitations of DLLME, such as changes in the volume of the sedimented phase and/or in the efficiency of the extraction, depending on the complexity of the original sample. On the basis of the results obtained in this research, it is expected that the sequential use of SPE and DLLME will broaden the applicability of the latter technique to complex liquid matrices, such as beverages, spirits, juices and even biological fluids. Acknowledgments This study has been supported by Spanish Government, Xunta de Galicia, and E.U. FEDER funds (projects PGIDIT06PXIB237039PR and PGIDIT06RAG50501PR-2). R. M. and M. R. thank the Spanish Ministry of Science and Innovation and the Xunta de Galicia for their FPU and IPP contracts, respectively. References 1. R. Rial Otero, B. Cancho Grande, J. Simal Gándara, J. Chromatogr. A 992 (2003) 121. 2. P. Cabras, A. Angioni, J. Agric. Food Chem. 48 (2000) 967. 3. P. Cabras, A. Angioni, V.L. Garau, F.M. Pirisi, J. Espinoza, A. Mendoza, F. Cabitza, M. Pala, V. Brandolini, J. Agric. Food Chem. 46 (1998) 3249. 4. S. Navarro, J. Oliva, A. Barba, G. Navarro, M.A. Garcia, M. Zamorano, J. Agric. Food Chem. 48 (2000) 3537. 5. European Food Safety Authority Scientific Report 227 (2009) 1-26. 6. J.J. Jiménez, J.L. Bernal, M.J. del Nozal, J. Bernal, L. Toribio, Food Chem. 104 (2007) 216. 7. M. J. Nozal, J.L. Bernal, J.J. Jiménez, M.T. Martín, J. Bernal, J. Chromatogr. A 1076 (2005) 90. 8. S. Navarro, A. Barba, G. Navarro, N. Vela, J. Oliva, J. Chromatogr. A 882 (2000) 221. 9. J. W. Wong, M.G. Webster, C.A. Halverson, M.J. Hengel, K.K. Ngim, S.E. Ebeler, J. Agric. Food Chem. 51 (2003) 1148. 10. M. Correia, C. Delerue-Matos, A. Alves, Fresenius J. Anal. Chem. 369 (2001) 647. 11. C.G. Zambonin, A. Cilenti, F. Palmisano, J. Chromatogr. A 967 (2002) 255. 12. L. Urruty, M. Montury, M. Braci, J. Fournier, J. M. Dournel, J. Agric. Food Chem. 45 (1997) 1519. 13. S. Millán, M.C. Sampedro, N. Unceta, M.A. Goicolea, E. Rodríguez, R.J. Barrio, J. Chromatogr. A 995 (2003) 135. 14. L.M. Ravelo-Pérez, J. Hernández-Borges, T.M. Borges-Miquel, M.A. Rodríguez-Delgado, Food Chem. 111 (2008) 764. 15. L.M. Ravelo-Pérez, J. Hernández-Borges, T.M. Borges-Miquel, M.A. Rodríguez-Delgado, Electrophoresis 28 (2007) 4072. 16. L. Urruty, M. Montury, J. Agric. Food Chem. 44 (1996) 3871. 17. D. A. Lambropoulou, I. K. Konstantinou, T. A. Albanis, J. Chromatogr. A 893 (2000) 143. 18. R. Rial Otero, C. Yagüe Ruiz, B. Cancho Grande, J. Simal Gándara, J. Chromatogr. A 942 (2002) 41. 19. P. Viñas, N. Aguinaga, N. Campillo, M. Hernández-Córdoba, J. Chromatogr. A 1194 (2008) 178. 20. E. Psillakis, N. Kalogerakis, Trends Anal. Chem. 22 (2003) 565. 21. Y. Liu, E. Zhao, Z. Zhou, Anal. Lett. 39 (2006) 2333. 22. M. Rezaee, Y. Assadi, M.-R. M. Hosseini, E. Aghaee, F. Ahmadi, S. Berijani, J. Chromatogr. A 1116 (2006) 1. 23. R.R. Kozani, Y. Assadi, F. Shemirani, M.R.M. Hosseini, M.R. Jamali, Talanta 72 (2007) 387. 24. M.A. Farajzadeh, M. Bahram, J.A. Jönsson, Anal. Chim. Acta 591 (2007) 69. 25. R.S. Zhao, C.P. Diao, X. Wang, T. Jiang, J.P. Yuan, Anal. Bioanal. Chem. 391 (2008) 2915. 26. M. García-López, I. Rodríguez, R. Cela, J. Chromatogr. A 1166 (2007) 9. 27. N. Fattahi, S. Samadi, Y. Assadi, M.R.M. Hosseini, J. Chromatogr. A 1169 (2007) 63. 28. E. Zhao, W. Zhao, L. Han, S. Jian, Z. Zhou, J. Chromatogr. A 1175 (2007) 137. 29. L. Fariña, E. Boido, F. Carrau, E. Dellacassa, J. Chromatogr. A 1157 (2007) 46. 30. N.L. Wolfe, R.G. Zepp, J.C. Doster, R.C. Hollis, J. Agric. Food Chem. 24 (1976) 1041.