ANALYST, JANUARY 1992, VOL. 117 67 Development and Evaluation of Analytical Methodology for the Determination of Aflatoxins in Palm Kernels Sadat Nawaz School of Biological and Chemical Sciences, Thames Polytechnic, Wellington Street, London SE 18 6PF, UK Raymond D. Coker" Mycotoxins Section, Natural Resources Institute, Chatham Maritime, Central Avenue, Chatham, Kent ME4 4TB, UK Stephen J. Haswell School of Chemistry, The University of Hull, Hull HU6 7RX, UK A rapid, simple and reproducible method for the simultaneous estimation of aflatoxins AFB1, AFB2, AFG1 and AFG2 in palm kernel samples has been developed by optimizing the sample preparation, solvent extraction, sample clean-up and quantification procedures. The aflatoxins are extracted from a slurried palm kernel sample with an acetone-water (80 + 20, v/v) mixture and the crude extract is cleaned up by solid-phase extraction using a phenyl bonded phase cartridge.The extract is passed through the cartridge with a water-methanol (93 + 7) mixture. Subsequent elution of the aflatoxins retained on the cartridge is achieved with a 3 m l aliquot of chloroform. The aflatoxin content of the eluate is quantified using a bi-directional high-performance thin-layer chromatography procedure. A critical evaluation of the proposed method was carried out by statistical comparison with the British Standard Method. The proposed procedure was shown t o be more efficient and precise. Consistent recoveries of over 90% were achieved from spiked palm kernel extracts and detection limits were found t o be 3.7, 2.5, 3.0 and 1.3 pg kg-1 for AFB1, AFB2, AFG, and AFG2 aflatoxins, respectively. Keywords: A flatoxin; palm kernel; sample preparation; solid-phase extraction; high-performance thin-layer chromatography Aflatoxins are highly toxic secondary metabolites of the moulds Aspergiffus ffavus and A.parusiticus. The moulds thrive on substrates rich in carbohydrates and lipids under the conditions of high temperature and relative humidity' that occur in tropical countries. The oil palm Efueis guineensis Jucquin3 exists in wild, semi- wild and cultivated states in the equatorial land areas of Africa, South East Asia and America. The main product of the palm is the fruit, which is rich in palm oil. The fruit also contains a nut, which is cracked in order to obtain the palm kernel.Palm kernels, which on average form 20% by mass of the whole fruit, are typically composed of at least 50% oil. Palm kernel oil is used mainly in the food and detergent industries. Palm kernel cake and meal, by-products of the palm kernel oil extraction process, are used as protein supplements in compound animal feeds.4 These tropical areas provide ideal conditions for mould growth and the production of aflatoxins on palm kernels5 both at the pre- and post-harvest stages. Damage caused to the kernels due to the mechanical cracking and insect infestation in addition to poor storage conditions can aid mould infection. The introduction of legislation to limit the levels of aflatoxins in foods and feedstuffs in over 56 countries6 necessitates the development of efficient analytical methods for quality control purposes.The British Standard Method (BS 5766) for the determina- tion of aflatoxins in palm kernels7 is a slow and costly procedure involving a chromatographic clean-up stage using a silica gel based column with quantification by thin-layer chromatography (TLC). A method using a combination of an immunoaffinity clean-up stage and quantification by high- performance liquid chromatography (HPLC) has also report- edly been applied to the analysis of palm kernels.8 However, no data describing the efficiency of the method have been *: To whom correspondence should be addressed. published. No other methods have been reported to date for the determination of aflatoxins in palm kernels apart from the methods described above.Consequently, a demand has arisen for a simple, rapid, accurate and reproducible assay method for the determination of aflatoxins in palm kernels. Solid-phase extraction procedures, which are ideally suited to automation, have greatly simplified the sample clean-up stage.Y The use of non-polar phenyl (PH) bonded phase cartridges for the clean-up of maize10 and peanut butter" extracts with subsequent quantification using high-perfor- mance thin-layer chromatography (HPTLC)l* has been re- ported. In this paper, the development and validationl3-15 of a method suitable for the determination of aflatoxins in palm kernel samples is described. It should be noted that, owing to the highly toxic nature of aflatoxins, extreme care16 must be taken when carrying out the procedures described in this paper.Experiment a1 Apparatus An Apex knife mill (3 mm screen) was used for grinding the kernels and a rotary cascade divider (Pascall Engineering, Crawley, UK) was employed for sub-sampling the ground samples. Waring blenders (Dynamics Corporation of America, New Hartford, C?', USA) were used for slurry preparation (4 I) and for the acetone-water extraction (1 1). Whatman No. 1 filter-papers were used for the filtration of blended mixtures. The clean-up apparatus, supplied by Jones Chromato- graphy (Hengoed, Mid-Glamorgan, UK), consisted of a vacuum manifold (Vacelut, A 1600) used in conjunction with a disposable 500 mg PH bonded phase cartridge (Cat. No. PH; 608303) selectively coupled with 4 ml (Cat. No. 600400), 25 ml (Cat.No. 602500) or 75 ml (Cat. No. 607500) reservoirs using68 ANALYST, JANUARY 1992, VOL. 117 1 2 75 ml reservoir - Phenyl cartridge 25 ml reservoir Phenyl cartridge Anhydrous sodium sulfate column Fig. 1 Palm kernel extract clean-up using bonded phase cartridges prior to HPTLC quantification. 1. Cartridge activation: pass 1 ml of methanol followed by 10 ml of water through the cartridge. 2, Extract clean-up: (i) 5 ml of aqueous acetone treated with 1 ml of lead acetate solution. (Addition of 1 g of Celite also aids the clean-up process.) (ii) Pass 5 ml of the treated aqueous acetone sample extract through the cartridge together with 5 ml of methanol and about 63 ml of water. under vacuum, at a rate of 1 0 ml min-1. 3 , Washing the cartridge: wash with 1 0 ml of water and dry the cartridge (2 min).4. Aflatoxin elution: with 3 ml of chloroform adapters (Cat. No. 636001) and Luer stopcock (Cat. No. A 16078) (Fig. 1). The glass vials (8 ml) used for eluate collection and work-up were supplied by Merck (Cat. No. 215/0073/05). Additional equipment used for the British Standard Method included a wrist action shaker (Voss Instruments, Maldon, Essex, UK), glass columns (22 x 300 mm), and a Buchi rotary evaporator (Laboratorium-Technik AG CH- 9230, Flawil/Schweiz, Switzerland). A sample concentrator (Tecam, UK, DM-Block DB-3) was used for drying chloroform extracts prior to quantification. A Perkin-Elmer Lamda 3 ultraviolet-visible (UV/VIS) spectro- photometer was used to determine the concentrations of the aflatoxin standard and spiking solutions.The aluminium backed HPTLC plates were supplied by Merck (Cat. No. 5547). The plates were spotted using an automated TLC sampler (Camag Cat. No. 27200). A conventional TLC tank and a continuous linear TLC tank were employed during the chromatographic development of the plates. 1' The TLC scanner I1 (Cat. No. 76610) and TLC integrator SP 4270 (Cat. No. 76650) (Camag, Switzerland) controlled by a personal computer with link-up software (Quadrant Scientific, UK) were employed for the densitometry measurements. 17 Reagents Lead acetate (20%) solution was prepared as described by Stoloff.18 All chemicals and solvents were AnalaR or HPLC grade (Merck) and distilled water was used throughout. Aflatoxin Standards Crystalline aflatoxin standards purchased from Sigma were diluted to concentrations of about 10 pg ml-1 in a mixture of benzene-acetonitrile (93 + 2) and the exact concentrations of the solutions were determined by UV absorbance measure- ments.l y Standard solutions, prepared by diluting the above solutions to give aflatoxin concentrations of 1.0 pg ml-1 (AFB,, AFGl) and 0.5 pgml-1 (AFB?, AFG?) in benzene- acetonitrile (98 + 2), were stored at -20 "C. [Caution: Aflotoxins are carcinogenic to humans. Gloves and other protective clothing must be worn for all operations involving the handling of these compounds or their solutions. All work should be carried out in a well ventilated area.161 Palm kernel samples collected from West Africa during the autumn of 1988 were used during this investigation.Procedure Sample preparation Riffle division. A finely ground palm kerncl sample (1 kg) was divided into six sub-samples, using a rotary cascade sample divider (spinning riffle). The aflatoxin content of each sub-sample was determined in replicate ( n = 4) using the proposed method. Comparison of methods used in sample dillision. A finely ground palm kernel sample (1.8 kg) was divided into two parts using a rotary cascade divider. One of the 0.9 kg sub-samples was slurried with 1.35 1 of water (1 : 1.5 m/v) in a 4 1 blender at high speed for 3 min. Ten aliquots (100s) of the resultant slurry were individually blended at high speed in a 1 1 blender with 240 ml of acetone for a further 3 min. Each of the ten mixtures was then filtered through a Whatman No.1 filter- paper. The aflatoxin content of each filtrate was determined using the proposed method. The other 0.9 kg sub-sample was divided into eighteen 50 g sub-samples using a rotary cascade divider. Ten of the sub-samples were individually blended with 300 ml of acetone and 75 ml of water in a 1 I blender and the resultant mixtures were filtered and analysed by solid-phase extraction and bi-directional HPTLC. Optimization of the solvent extraction procedure Choice of solvent. Sets of five aliquots (100 g) of palm kernel slurry (1 : 1.5 m/v) were blended in a 1 1 Waring blender, with 240 ml portions of five different acetone-methanol mixtures (Table 3). The ratio of water to organic phase in the solvent mixture remained constant at 20 + 80. The resultant mixtures were analysed by solid-phase extraction and bi-directional HPTLC.Optimization o f the soliient : sample ratio. Sets of five aliquots of palm kernel slurry were extracted with mixtures of acetone-water with solvent : sample ratios which ranged from 7.5 : 1 to 25 : 1 (Table 4). Effectirieness of the extraction procedure. Replicate 100 g palm kernel slurries (60 ml of water : 40 g of sample) were each blended with 240ml of acetone and filtered through a Whatman No. 1 filter-paper. The first 230ml of the filtrate were collected in a measuring cylinder and taken through the clean-up and quantification procedures. The residue (which contained 70 ml of solvent mixture) and the filter-paper were transferred into a Biichner funnel and washed with excess (approximately 480 ml) of acetone-water (80 + 20), the wash was made up to a final volume of 500 ml and quantified for aflatoxins.The residue in the Buchner flask was then blended with 250ml of acetone-water (80 + 20), filtered, cleaned-up and again quantified for aflatoxins. Use of lead acetate for the clean-rip The minimum amount of lead acetate required to achieve a sufficient precipitation of the colloidal components present in the crude acetone-water (80 + 20) extracts was investigated. Crude sample extracts were mixed with various volumes (0-5 ml) of a 20% m/v lead acetate solution, followed by the proposed clean-up and quantification steps. The process was duplicated for each volume of lead acetate. Composition of the mobile phase Use of acetic acid. Aliquots ( 5 ml) of palm kernel extracts were applied to pH bonded phase cartridges along withANALYST, JANUARY 1992, VOL.117 69 aliquots (63 ml) of acetic acid solutions (0-3% v/v) and quantified using the proposed method. The process was duplicated for each acetic acid concentration. Volume of methanol in the mobile phase. Various volumes of methanol and the crude acetone-water extract were applied to the PH cartridge along with various volumes of distilled water, and the cleaned-up extracts were quantified using bi-directional HPTLC. Elution and work-up stage Two experiments were conducted in order to optimize the elution and work-up stages. Firstly, 4.8 ml of a mixed aflatoxin solution in benzene-acetonitrile (98 + 2), was evaporated under nitrogen at 45 "C on a sample concentrator.The residue was dissolved in 120 ml of an acetone-water mixture (80 + 20) and homogenized. Twenty four aliquots ( 5 ml) of the above solution were each treated by the PH bonded phase clean-up procedure. Replicate elutions ( n = 4) of the aflatoxins were carried out with varying amounts of chloroform (1, 2, 3, 4, 5 and 7 ml). The chloroform extracts were dried and quantified, for the aflatoxins, using bi-directional HPTLC. In the second experiment, 42 equal volumes (200~1) of a mixed aflatoxin standard solution in benzene-acetonitrile (98 + 2) were evaporated to dryness. The residues were reconsti- tuted in seven different volumes of chloroform (1-7 ml) to give six replicates at each volume. Solvent evaporation was performed under a gentle stream of nitrogen at 45°C on a sample concentrator.The times required to evaporate increas- ing volumes of chloroform are summarized in Table 5. Aflatoxin quantification was carried out using bi-directional HPTLC. Comparison of Analytical Methods Two 1 .0 kg samples of naturally contaminated (approximately 100 and 600 pg kg-* of AFB!) ground palm kernels were halved using a rotatory cascade divider, to produce two representative sub-samples (500 g), [IA and IB (100 v g kg-1 AFBI) and IIA and IIB (600pgkg-1 AFB1)] at each contamination level. British Standard Method7 Sub-samples IA and IIA were each divided into ten equal 50 g portions. Each portion was shaken with 25 ml of water, 250 ml of chloroform and 25g of diatomaceous earth in a 500ml stoppered flask, on a wrist action shaker, for 30min.The mixture was passed through a Whatman No. 1 filter-paper and the filtrates were combined to give bulked extracts, one at each level of contamination. The following procedure was repeated ten times for each of the bulked extracts. A chromatographic column was prepared by two thirds filling a glass column (22 x 300mm) with chloroform, to which anhydrous granular sodium sulphate ( 5 g) was added, followed by the addition of silica gel (log) as a chloroform slurry. After the silica gel had settled, a further 10 g of sodium sulfate were added. The chloroform was allowed to drain until it just reached the upper surface of the sodium sulfate layer. A 50 ml portion of the bulked extract was mixed with 100 ml of hexane in a 250 ml measuring cylinder and the solution was quantitatively transferred into the chromatographic column.The solution was allowed to pass through the column at a rate of about 10 ml min-1 until it was level with the upper surface of the sodium sulfate layer. The column tap was then closed and 100ml of anhydrous diethyl ether were added to the column. The tap was then reopened and the eluate allowed to flow until it was again level with the upper surface of the sodium sulfate layer. The aflatoxins were eluted with 150 ml of a chloroform-methanol mixture (50 + 50), the eluate was collected in a 500 ml round-bottomed flask and evaporated to dryness on a rotary evaporator. The residue was quantitatively transferred into an 8 ml glass vial with a small amount ( 5 ml) of chloroform. Proposed bonded phase clean-up method The sub-samples IB and IIB (500g) were each slurried by blending at high speed €or 3 min, with 750 ml of water in a 4 I Waring blender.A 100 g portion, from each of the two slurries, was extracted by blending with 240 ml of acetone in a 1 1 Waring blender and filtered through a Whatman No. 1 filter-paper. An aliquot ( 5 ml) of the crude extract was mixed with methanol ( 5 ml), 20% m/v lead acetate solution (1 ml), Celite (1 g) and water (63ml). The mixture was then applied to a solvated PH bonded phase cartridge at a rate of 10ml min-I. (The solvation step involved the passage of 10ml of methanol followed by 10ml of water through the cartridge.) After washing the cartridge with 10 ml of water, the aflatoxins were eluted using 3ml of chloroform.The eluate was passed through a 4 ml reservoir containing anhydrous sodium sulfate, to eliminate any water, and collected in an 8 ml glass vial. The above procedure was carried out under reduced pressure on a Vacelut vacuum manifold. Work-up. Each chloroform solution was evaporated under a steady stream of nitrogen at 45 "C on a sample concentrator in order to give clean, dry sample extracts. Quantification using bi-directional HPTLC. High-perfor- mance TLC plates (20 x 20 cm) were cut in half and immersed in methanol for 1 h in order to remove atmospheric contami- nants deposited on the plate during storage. The plates were dried for 5 min in an oven at 100°C, situated in a fume cupboard, and stored in a desiccator until required. The following procedures were performed under darkened conditions.The clean, dried sample extracts were dissolved in 300 pI of a benzene-acetonitrile mixture (98 + 2). Aliquots of the above extracts ( 5 PI) along with external standards (1 p1 aliquots of a mixed aflatoxin solution) were applied, using an autosampler, as a row of spots at 5 mm intervals 3 cm from the top edge of an HPTLC plate. One 10 x 20 cm HPTLC plate accommodated up to 30 sample spots together with three standards along the 20 cm edge. A strip of silica gel (approximately 3mm wide) was removed from the edges of the plate, parallel to the direction of development, in order to eliminate edge effects. A further sample clean-up was performed by developing the plate in 20 ml of diethyl ether for 17 min in a continuous horizontal tank.[NB. The eluent reached the top of the plate within 3-4 min. Continuous development was then allowed to proceed for a further 13-14 min, hence there is no solvent front.] The plate was dried for 3min in a darkened chamber under a stream of nitrogen and the top portion (2cm) of the plate, containing sample interferences, was cut and removed using a sharp knife. The plate was rotated through 180" and developed for 20min in a conventional TLC tank, using 20ml of a chloroform-xylene-acetone (CXA) (6 + 3 + 1) mixture for 20 min. After drying ( 5 min) the bottom portion (1 cm) of the plate was removed, as before, in order to decrease the development time and improve the resolution of the aflatoxin spots. This was followed by another CXA development (16 min).These times for the CXA development were found to be optimum for the conditions existing during the method development. Factors such as temperature variations may have a slight effect on the method and should be modified accordingly until the solvent front is approximately 1 cm from the top of the plate. The plate was finally dried (1 min) in a fan assisted oven at 100 "C before quantification of the aflatoxins by UV fluorescence in the reflectance mode using the TLC I1 scanner supported by appropriate software. 17 Statistical Evaluation of the Proposed Method Nine 30ml aliquots of aflatoxin-free crude palm kernel extracts and nine 30 ml aliquots of an acetone-water (80 + 20)70 ANALYST, JANUARY 1992, VOL. 117 mixture were each accurately spiked with an acetone-water (80 + 20) solution of aflatoxins, in order to produce two sets of nine concentrations in the range 0-300 pg kg-I for AFBl and AFG1, and 0-150pgkg-1 for AFB2 and AFG?.Aliquots (5m1, in replicates of six) from each of the spiked crude extracts and solvent mixtures were processed using the proposed clean-up method followed by quantification using HPTLC. Nine solutions of aflatoxins in benzene-acetonitrile (98 + 2) within the same concentration range of 0-300 pg kg-1 for AFRl and AFG, and 0-150 pg kg-l for AFB? and AFG2 were also quantified by bi-directional HPTLC. Calculation Effective weight The effective weight of the extracts from the proposed clean-up procedure were calculated using the following equation: Vextract X Msample (Mslurry - Msampie) + Vsolvent) Effective weight = where Vextract = volume of extract = 5 ml; Msample = mass of sample in the slurry; Mslurry = mass of slurry = 1OOg; and Vsolvent = volume of the organic solvent = 240 ml.The mass of sample in the slurry is given by mass of sample slurried mass of the total slurry X Mslurry Msampie = If the mass of sample slurried = lOOOg, and the mass of the total slurry = 2500g then Msampie = 100 x 1000/2500 = 40g Hence, 5 x 40 Effective weight = = 0.667 g (100 - 40) + 240 Spiking volume The volumes of the spiking solution required to produce an artificially contaminated palm kernel extract at a specified level was calculated using the following equation D x effective weight V = C where V = volume of the spiking solution required (PI); D = desired contamination level (pg kg-I); and c = concentration of the spiking solution (pg ml-1). Results and Discussion Method Development The results discussed below refer to AFBl only (unless otherwise stated), as AFBl was the major toxin found in naturally contaminated samples in this study.Sub-sampling and sample preparation It was important to establish early in this work whether the sub-sampling of ground palm kernels would introduce bias into the analytical procedures. The sub-sampling, using a rotary cascade divider, and the analysis of a 1 kg sample of ground palm kernels facilitated the separation of variances due to both the sub-sampling and analytical procedures. From the analysis of variance (ANOVA) calculations, these variances were found to be 4.67 and 81.728, respectively (Table 1).The calculated value of F was found to be 1.229, compared with the critical value for Fs,18 (p = 0.05) of 2.773, thus confirming that the variance due to sub-sampling did not significantly differ from zero. Table 1 Analysis of variance for AFBl to estimate the variances associated with spinning riffle sub-division. Sum Degrees Variance of of Mean com- Between sub-samples 502.04 5 100.41 4.67 Within sub-samples 1471.10 18 81.73 81.73 Source of variation squares freedom squares ponent - - Total 1973.14 23 Table 2 Comparison of sample preparation techniques Dry sampling Slurry Parameter method method Mean recovery of AFB ,/pg kg- 1 120.5 207.2 Variance 105.4 340.0 Standard deviatiodpg kg- 1 10.3 18.4 Relative standard deviation (Yo) 8.9 8.5 No.of replicates, n 10 10 Computed t statistic = 13.0 Ratio of variances = 3.2 F9,Y 0, = 0.95) = 4.03 t 0, = 0.05) = 2.1 Table 3 Comparison of solvent systems for the extraction of aflatoxins Composition of acetone- methanol- water (%) a, 80+00+20 b, 70 + 10 + 20 c, 60 + 20 + 20 d, 40 + 40 + 20 e. 00 + 80 + 20 * n = 5 . Average AFB I extracted*/ - Yg kg-I 475.2 441 .5 433.8 409.2 360.7 95% confidence interval*/yg kg-1 Upper Lower 492.6 457.4 465.5 417.5 456.1 41 1.5 420.1 398.3 394.7 326.7 Standard deviation/ 14.0 19.3 18.0 8.8 27.4 P8 kg- I The production of sub-samples using a rotary cascade divider (spinning riffle) was compared with the slurry method”) by preparing two equivalent 0.9 kg samples of ground palm kernels according to the two sample preparation methods.Analytical results obtained using each of the sample preparation methods showed good repeatability, with relative standard deviations of 8.9 and 8.5% for the spinning riffle and slurry techniques, respectively. Statistical analysis of the results (Table 2) showed that there was no significant difference at the 5% significance level between variances associated with the methods. However, the slurry technique was found to extract significantly more aflatoxins than the dry extraction method. Optimization of extraction procedure f o r aflatoxins Choice of solvent. Solvent systems composed of acetone- methanol-water in various ratios were assessed for their abilities to extract aflatoxins from palm kernel samples. Inspection of the confidence intervals (Table 3) for the mean recoveries of AFB, indicated that solvent mixture ‘a’ ex- tracted significantly more AFBl than the other solvent systems with the exception of solvent system ‘b’.However the t-test used to compare the mean extracted value of AFB using the solvent systems a and b ( t = 3.16 and critical value for t (p = 0.05) for 8 degrees of freedom = 2.31), showed that the solvent system a extracted significantly more of AFBl than system b at the 95% level of confidence. An F-test showed that the variance associated with the solvent system a did not differ significantly from the variances for the other extraction mixtures. Optimization of tfie extraction solvent volume. Slurried palm kernel samples were extracted with varying volumes of the acetone-water mixture in order to optimize the sample : sol- vent ratio.The recovery results for AFB, (Table 4) indicated that changing the ratios did not lead to a pronounced change inANALYST, JANUARY 1992, VOL. 117 71 Table 4 Volume of acetone-water (80 + 20) mixture required for the efficient recovery of aflatoxin Amount of slurry/g 100 75 60 50 35 25 * n = 5 . Volume of water added/ml 0 9 21.6 30 35 35 Volume of acetone added/ml 240 2 16 230.4 240 224 200 Total solvent volume/ml 300 270 288 300 280 250 Solvent: sample ratio 7.5: 1 9 : 1 12: 1 15: 1 20: 1 25 : 1 Effective weight/g 0.67 0.56 0.42 0.33 0.25 0.20 95% confidence interval for AFB, extracted*/yg kg-I Upper Lower 320.3 277.6 326.5 275.6 332.9 268.8 308.2 286.3 291.2 254.4 297.4 250.9 the quantity of aflatoxins extracted from the palm kernel samples.The lowest possible solvent : sample ratio of 7.5 : 1 v/m was adopted as it afforded the highest effective weight and, hence, lower detection limits. Consequently, the extraction of a 100 g slurry containing 40 g of sample and 60 ml of water, with 240 ml of acetone, was adopted as the optimum extraction procedure. Effectiveness of the extraction procedure. The investigation showed that the first 230 mi of filtrate collected contained 2.03 pg of AFB1. The residue and the filter-paper were washed with an excess of the solvent and contained 0.61 pg of AFB1. Quantification of the third extract failed to show the presence of any aflatoxin. From these results, it would appear that the initial extraction with 300 ml of acetone-water (80 + 20) solvent mixture removed all the aflatoxins from the matrix.However, owing to the difficulty in recovering all of the solvent approximately 70 ml (23.33%) remained unfil- tered. The concentrations of aflatoxins (8.83 and 8.71 pgml-1) in these two fractions were not found to be different, using a t-test, at the 95% level of significance, indicating that no gradient was present in the two portions of the solvent mixture analysed. Precipitation of interfering components prior to column clean-up Lead acetate18 solution has previously been used to eliminate, by precipitation, proteins, lipids and other colloidal com- ponents from cottonseed extracts. However, because of the toxicity of lead acetate, studies were carried out in order to minimize its usage.Visual inspection of the extracts obtained from the bonded-phase clean-up, showed that the use of low levels of lead acetate afforded oily residues. This was particularly apparent when no lead acetate was used. The presence of the oily residues artificially increased the volume of the cleaned-up benzene-acetontrile extract, causing a dilution effect and an increase in interferences, which led to lower recoveries. As the volume of lead acetate was increased above 1 mi, no differences were observed in the aflatoxin recoveries or the cleanliness of the extracts. Hence, it was concluded that 1 ml of lead acetate solution would be sufficient for the satisfactory precipitation of interfering colloidal components from palm kernel extracts. Celite washed with methanol was also used as a filter aid.composition of the mobile phase In a previous method reported for the analysis of maize,") the mobile phase was composed of a methanol extract in a 1% acetic acid solution. In these studies, the presence of acetic acid in the mobile phase appeared to have no effect on the aflatoxin retaining properties of the PH bonded phase cartridges. More importantly, however, a visual comparison of the resultant extracts showed that the acid-free mobile phase led to the cleanest extract. The presence of methanol has been reported to have a significant effect on the retention of aflatoxins o n the PH I 100 90 - 8 H 8o > 8 70 a, U 60 50 0 2 4 6 Volume of chloroform/ml Fig. 2 Effect of changing the volume of chloroform as eluting solvent on the percentage recoveries of aflatoxins.A, AFB,; B, AFB2; C, AFG,; and D, AFG2 Table 5 Effect of drying times on aflatoxin recoveries Volume of chloro- form/ml 1 2 3 4 5 6 7 * n = 6 . Average drying ti me/m i n 8 13 18 23 27 32 37 Average recovery* 99.6 93.5 93.8 81.7 82.6 82.2 73.2 ( Y o ) 95% confidence interval* Upper 102.2 96.5 96.9 83.5 86.0 84.1 75.2 Lower 97.0 90.5 90.7 79.9 79.2 80.3 71.2 cartridges.?' This was confirmed in the present study as a better retention of the aflatoxins contained in the acetone- water extracts was achieved when methanol was added to the extracts. The volume of methanol required was found to be equal to the volume of acetone in the mobile phase. However, as the proportion of the organic phase (acetone and methanol) was increased to above 20%, the recoveries of aflatoxin decreased.It was concluded, therefore, that the mobile phase should be composed of 5ml of acetone-water extract, 5ml of methanol, 1 ml of lead acetate solution (20%) and 63 mi of distilled water. Elution and work-up stage Two experiments were performed to optimize the elution and work-up stages. The fifst experiment, conducted to optimize the volumes of chloroform required for the maximum elution, showed that the recoveries of aflatoxin initially increased with an increase in the elution volume (Fig. 2). However, as the volume of the chloroform approached the 2ml level, the recoveries levelled off and gradually dropped as the volume of72 100 90 8 80 ' 70 h v B 8 a 60 chloroform was increased above 4 ml. This apparent loss of aflatoxins was investigated further in the second experiment.The analyses of variance [Fratio = 84.3, F6,35 (p = 0.05) = 2.41 of the results from the second experiment (Table 5) showed that as the exposure to heat increased during evaporation the recovery of aflatoxin was affected signifi- cantly. Fig. 3 also shows that the recovery of aflatoxin decreased as the increasing volumes of chloroform required a longer exposure to heat during the evaporation step. It is evident from Figs. 2 and 3 that the 'B' aflatoxins were less susceptible to losses during the elution and the work-up stage than the 'G' aflatoxins. Furthermore, AFB2 showed a higher recovery than AFBl and, similarly, AFG2 showed a better recovery than AFG1, giving recoveries in the order AFB2 > AFBl > AFG2 > AFGl .The same order of recovery from spiked acetone-water mixtures was found to exist in a later evaluation study (Table 6). It may also be postulated that an increase in the volume (and surface area) of the chloroform solution will lead to a greater irreversible adsorption of aflatoxin onto the wall of the glass vial. - - - - - Quantification The bi-directional HPTLC method was found to be both precise and accurate. The TLC scanner was calibrated for each plate using external standards to minimize inter-plate variabil- ity. Typical RF values, when using the bi-directional HPTLC 40 50 I 0 2 4 6 Volume of chloroform/ml Fig. 3 Effect of exposure to heat on aflatoxin recovery. A, AFB,; B, AFB,; C, AFG,; and D, AFGz ANALYST, JANUARY 1992, VOL.117 described above, were found to be 0.28, 0.3, 0.4 and 0.45 for AFBI, AFB2, AFGl and AFG2, respectively. Detection limits of 18.9, 6.4, 15.1 and 8.4pg were calculated (Table 6) for AFB,, AFB2, AFGl and AFG2, respectively, using this method. Method Validation Statistical evaluation of the proposed method The recovery data from the three evaluation procedures were used to construct calibration lines for aflatoxins using a weighted regression method.14 These statistical results were used to calculate the detection limits15 and percentage recoveries and to detect the existence of any inherent relative and/or systematic errors for the AFB1, AFB2, AFGl and AFG2. Calibration data, summarized in Table 6, for the four aflatoxins were obtained using recovery results from: (1) HPTLC quantification of the aflatoxins after a spiked palm kernel extract had been subjected to the proposed clean-up procedure; (2) HPTLC quantification of the aflatoxins after spiked acetone-water (80 + 20) solutions had been subjected to the proposed clean-up procedure; and (3) HPTLC quantifi- cation of various concentrations of the aflatoxins in a benzene-acetonitrile (98 + 2) solution.The presence of a systematic error was indicated when the 95% confidence limits for the regression intercept did not include the value of zero. Systematic errors were found to be present only for AFB2 and AFG2 standards (Table 6). Even for these, the lower limits of the lines passed very close to the origin (0.172 and 0.102). Relative errors were estimated by the percentage deviation of the confidence intervals of the regression slope, from an expected slope of unity.Such errors were found to exist for all the calibration lines (Table 6). The decreased aflatoxin recoveries, shown by the relative errors, can be attributed to the losses incurred and were discussed under Elution and work-up (Figs. 1 and 2). The overlap of the 95% confidence intervals of the regression lines, for the individual aflatoxins, indicated that the accuracy of the proposed method was not affected by the presence of sample interferences or by the solid-phase extraction step. This overlap existed for AFBI, AFB2 and AFG?. However, the regression lines for AFGl did not completely overlap as the recoveries from the spiked acetone- water (80 + 20) mixture were significantly lower than from the spiked extracts. This effect was caused by the presence of interferences around the AFGl spot on the HPTLC plate, Table 6 Summary of the calibration data for aflatoxins 95% confidence limits y-Interceptlpg kg- Slope of the line Detection limit Correlation Relative error Sample Lower Upper Lower Upper ELg kg- ' Pi? coefficient (% range) - 1.595 -0.734 -0.425 -0.068 -0.025 0.172 - 1.931 -0.295 -0.328 -0.504 -0.275 0.102 0.813 1.274 0.190 1.004 1.154 0.440 0.615 1.482 0.824 0.488 0.327 0.529 0.893 0.943 0.922 0.92 0.92 0.962 0.975 0.847 0.946 0.887 0.92 1 0.948 0.980 1.002 0.991 0.981 0.971 1 .oo4 1.073 0.911 1.031 0.964 0.985 0.997 3.74 2.38 1.70 2.48 1.04 0.58 3.04 0.98 1.36 1.26 0.73 0.76 41.6 26.5 18.9 27.5 11.6 6.4 33.8 10.9 15.1 14.1 8.2 8.4 0.9997 0.9996 0.9998 0.9982 0.9996 0.9999 0.9985 0.9998 0.9997 0.9968 0.9981 0.9998 - 10.7, -2.0 -5.7, 0.2 -7.8, -0.9 -8.0, - 1.9 -8.0, -2.9 -3.8, 0.4 -2.5.7.3 - 15.3. -8.9 -3.4, 3.1 -11.3. -3.6 -7.9, - 1.5 -5.2. -0.3 * Spiked extracts subjected to the proposed methodology. t Spiked solvent mixture (80 + 20, acetone-water) subjected to the proposed methodology. $ Standard aflatoxin solutions (98 + 2, benzene-acetonitrile), of equivalent concentration to other samples, quantified by bi-directional HPTLC.ANALYST, JANUARY 1992, VOL. 117 73 Table 7 Comparison of the precisions of the proposed method and the British Standard Method PH bonded phase method British Standard Method Degrees of Degrees of Degrees of Degrees of freedom X S2 freedom s2 freedom x S2 Sample freedom S' 1 9 13.6 122.8 9 83.6 752.7 2 9 866.7 7800.4 9 3065.8 27 592.5 - 28 345.2 Total 18 - 7923.2 18 s'p" 7923.2118 = 440.2 Spt = 21.0 Variance ratio (FObh) = 1774.7/440.2 = 3.58 * Pooled variance. -t Pooled standard deviation.28 345.2118 = 1474.7 = 39.7 Table 8 Amount of AFB, extracted using the proposed method and the British Standard Method 95% confidence Mean intervalslpg kg-' recoveries/ Sample Method pgkg-1 Lower Upper 1 PH-bonded 91.4 88.2 94.6 1 British Standard 63.4 60.4 66.6 2 PH-bonded 694.6 691.4 697.8 2 British Standard 512.0 478.3 515.2 which enhanced the fluorescence of this component. The presence of these interferences was confirmed by spraying the plate with a 50% v/v sulphuric acid solution.19 After spraying, the plate was dried in the fume cupboard and viewed under UV (365nm) light; the aflatoxin was seen to have changed fluorescence colour from green to yellow. However, the interfering compounds still displayed greenish fluorescence.This particular interference problem was only encountered during the validation studies, as the aflatoxin-free sample had been in store for a long period, leading to the development of interfering compounds not normally associated with palm kernels. However, the apparent loss of AFGl at the work-up stage (Figs. 1 and 2) was thought to be the main cause of the low over-all recovery of AFG experienced during this validation exercise. Comparison of methods The aflatoxin contents of two naturally contaminated palm kernel samples were quantified using the proposed method and the British Standard Method.' The variances associated with the two methods were compared using the pooled standard deviations procedure13 and the results are given in Table 7.The observed figure for F ratios (3.58) was found to be higher than the critical value [Fls,lx (p = 0.975) = 2.61 and thus it was concluded that the two methods had significantly different variances and that the PH bonded phase method was more precise than the British Standard Method. The abilities of the two methods to extract aflatoxins from the palm kernel samples were compared by setting up a table of 95% confidence intervalsl3 for the mean recoveries (Table 8). Table 8 clearly shows that there is no overlap in the confidence intervals for the two methods, indicating that the proposed method extracts significantly more of AFBl than the British Standard Method.Conclusion The proposed methodology for the extraction and determina- tion of aflatoxins in palm kernels has been found to be more effective than the British Standard Method in reducing time and solvent costs, affording improvements in reproducibility (average relative standard deviation of 4.1% compared with 12.6% for the British Standard Method), efficiency and accuracy (extracted 40% more of AFB,). These improve- ments were attributed to the semi-automation of the clean-up stage (using a twelve-place vacuum manifold), the commercial availability of the pre-packed PH cartridges and the use of automated HPTLC procedures (capable of 30 quantifications in a single process), The bi-directional HPTLC procedure facilitated the removal of persist en t interfering com pounds.The extracts were suited to the HPTLC quantification which utilizes the natural fluorescent properties of the aflatoxins. These factors contributed to the low detection limits for the proposed method, which were well below the current legis- lative levels.6 The proposed method gave excellent results over the ranges of aflatoxin concentrations investigated. The authors are grateful to the Natural Resources and Environmental Department of the Overseas Development Administration for funding this work. APPENDIX Summary of the Proposed Methodology The following method for the estimation of aflatoxins in palm kernel samples is proposed. ( a ) Grind the sample to afford a free-flowing product.(b) Sub-divide, using a spinning riffle, to afford a 1 kg (c) Slurry the 1 kg sub-sample with 1.5 I of water. ( d ) Blend 100g of the slurry with 240ml of acetone and filter the mixture. ( e ) Solvate a PH cartridge by passing 10ml of methanol followed by 10 ml of water through it under reduced pressure. v) Mix 5 ml of the filtrate with 5 ml of methanol, 1 ml of lead acetate solution (20% m/v), 1 g of Celite and 63 ml of water and pass the mixture through the solvated PH cartridge under reduced pressure at a rate of 10 ml min-1 (Fig. 1). (8) Wash the cartridge with 10ml of water and dry the cartridge for 2 min. ( h ) Attach the cartridge to a 3 ml reservoir containing 1 g of anhydrous sodium sulfate, and elute with 3 ml of chloroform.Collect the eluate in an 8 ml glass vial. (i) Dry the elute at 45 "C under a gentle stream of nitrogen and dissolve the residue in 300 pl of benzene-acetonitrile (98 (j) Apply 5 1-11 of the benzene-acetonitrile solution to an ( k ) Quantify using bi-directional HPTLC. sub-sample. + 2). aluminium-backed HPTLC plate. References I Hsieh, D. P. H.. in Food Toxicology. A Perspective on the Relative Risks, eds. Taylor, S . L . . and Seanlan, R. A.. Marcel Dekkcr, Ncw York, 1989. p. 1 1 .74 2 Cuero, R. G., Smith. J. E.. and Lacey, J.. Trans. Br. Mycol. Soc., 1987, 89, 221. 3 Hartley, C. W. S . . The Oil Palm, Longman Group. Harlow, 3rd edn., 1988. 4 Shibata. M., and Osman. A. H., JARQ, 1988, 22, 77. 5 Moore, A., in Fifth Meeting on Mycotoxins in Animal and Human Health, eds. Moss, M. O., and Frank, M., University of Surrey, 1984, 117. 6 Van Egmond, H. P., Food Addit. Contam.. 1989. 6, 139. 7 British Standard Method for Analysis Feeding Stuff. (IS0 6651-1987). BS 5766. Part 7. 1988. 8 Martin. C. N., Mulholland, F., and Garner, R. C., Feed Compounder, 1987. 7. 10. 9 Sorbent Extraction Technol. ed. Vanhorne, K. C., Analyti- chem. Harbor City, CA, 1985. 10 Tomlins, K . I . , Jewers, K., and Coker. R. D., Chromato- graphia, 1989. 27, 67. 11 Dell, M. P. K . , Haswell, S. J.. Roch, 0. G., Coker. R. D., Medlock, V. F. P., and Tomlins. K.. Analyst. 1990, 115, 1435. 12 Tornlins. K. I . , Jewers, K., Coker, R. D.. and Nagler. M. J., Chromatographia, 1989, 27. 49. 13 Wernimont. G. T.. in Use of Statistics to Develop and Evaluate Analytical Methods, ed. Spendlcy. W., Association of Official Analytical Chemists, Arlington, VA. 1985. 14 Miller, J. C.. and Miller, J. N.. Statistics for Analytical Chemistry, Ellis Horwood, Chichester. ANALYST, JANUARY 1992, VOL. 117 15 16 Analytical Methods Committee. Analyst. 1987, 112, 199. World Health Organization/International Agency for Rcscarch into Cancer eds. Castegnaro. M., Hunt. D. C., Sansone. E. B., Schuller, P. L . . . Siriwardana. M. E., Telling, G. M.. Van Egmond. H. P., and Walker. E. A., ]ARC Scientific publication No. 37, Lyon, France. 1980. Coker, R. D.. Jewers, K.. Tomlins. K. I . . and Blunden. G., Chr-omatographia, 1988. 25, 875. Stoloff, L., and Scott, P. M., in Official Methods of Analysis of the Association of Official Analytical Chemists. ed. Williams, S . , Association of Official Analytical Chemists. Arlington, VA, 1984. pp. 380-484. Cokcr, R. D., Jones, B. D., Nagler. M. J . , Gilman. G. A., Wallbridge, A. J., and Panigrahi, S . , in Mycotoxin Truining Manual, ODNRI, London, 1984, Section B:2. Velasco. J.. and Morris, S. L.. J . Agric. Food Chem.. 1976.24. 86. Bradburn, N., Coker, R. D., and Jewers. K.. Clrromatogr-aphia, 1990, 29, 177. 17 18 19 20 21 Paper 0104949B Received November 5, 1990 Accepted July 2, 1991