ANALYST, JANUARY 1993, VOL. I18 89 Determination of Malonaldehyde in Human Plasma: Elimination of Spectral Interferences in the 2-Thiobarbituric Acid Reaction Anunciacion Espinosa-Mansilla, Francisco Salinas and Amparo Rubio Leal Department of Analytical Chemistry, University of Extremadura, 06077 Badajoz, Spain A selective, derivative spectrophotometric method has been developed for the determination of malonal- dehyde (MLD), based on a reaction with 2-thiobarbituric acid (TBA). The proposed method has been applied t o the determination of MLD in human plasma. A study t o eliminate several spectral interferences is described. A comparative study of the results obtained using the proposed derivative method and a conventional TBA method applied t o human plasma is presented and the advantages of the proposed method over the conventional method for the determination of MLD in human plasma are evaluated.Keywords: Human plasma; malonaldeh yde determination; derivative spectrophotometry; 2-thiobarbituric acid test Malonaldehyde (MLD) is a volatile side product generated in the enzymic oxygenation of arachidonic acid and an end product of the oxidative degradation of lipids. The metabolic- ally uncoupled oxygenation of polyunsaturated fatty acid leading to lipid-derived MLD is a degenerative process of oils, foodstuffs and lipid-rich biomolecular assemblies (mem- branes, lipoproteins), which has been somewhat imprecisely termed auto-oxidation. 1 The ability of MLD to alter and/or cross-link a variety of biological molecules might contribute to its toxicity,2 and its mutagenic and/or carcinogenic properties could reflect adduct formation with nucleic acid bases.3.4 Covalent modification of lipoproteins with MLD might play a pathogenic role in atherosclerosis .s The major urinary metabolite of MLD in humans is 2-N-acetyl-6-N-( 1-formylvinyl)lysine, which appears to be derived from the degradation of proteins, phospholipids and nucleic acids, which have reacted with MLD via their amino groups.6.7 Stocks et a1.8 employed the MLD level as a measure of human red cell lipid auto-oxidation by H202 and their assay has become the standard (conventional method) for the determination of the susceptibility of red blood cells to lipid peroxidation. In its free form, MLD can be determined by ultraviolet (UV) absorptiome try ,Y polarography 1 0 and high-performance liquid chromatography (HPLC). 1 1.12 The 2-thiobarbituric acid (TBA) test is an extensively used procedure for determining MLD; one molecule of MLD reacts with two molecules of TBA with the elimination of two molecules of water.'3 This reaction produces a fluorescent red pigment with a high molar absorptivity, between 5- and 10-fold greater than that of MLD itself in the UV.1.' The colour intensity and its ready formation have prompted detailed investigation of the reaction between MLD and TBA.15.16 During the TBA reaction, many lipid-derived monofunc- tional aldehydes [2-furaldehyde (FIJR), 5-hydroxymethyl-2- furaldehyde (HMF), ethanal-sucrose (ETA-SUC) binary mixtures and glyoxal (GLY), etc.] form adducts with TBA which fluoresce and/or contribute, with a yellow or orange colour, to the reaction.17-20 Hence, these reactions can directly interfere with the spectrophotometric quantification of MLD.Another source of error arises from the presence of pigments, particularly in plant materials, contributing to spuriously high estimates of MLD content (e.g., biliverdin, BIL) .20 In addition, sample acidification can induce turbidity or precipitate non-lipid sample constituents (e.g., proteins), and clarification of the sample prior to the reaction is generally accomplished by centrifugation." Pigments can be easily extracted into organic solvents from the TBA-test mix- tures,22923 although no solvent extraction procedure yet reported is selective toward any particular pigment.In a recently published review,24 it was concluded that the determination of MLD and interpretation of sample MLD content and TBA response in studies of lipid peroxidation require caution, especially for biological systems. In human plasma, increased susceptibility to lipid peroxida- tion of the erythrocyte membrane has been described in haematological disorders, as well as in a b-lipoproteinaemia and tocopherol deficiency.25 Largilliere and Melanconl2 have found in normal plasma samples a concentration of 3.8 pmol 1 - 1 of material that reacts with TBA using the conventional method, and no detectable MLD content when HPLC is used. Hence, whether the elevated MLD content reported in diverse pathological conditions is genuinely due to an increased production of MLD or to other unknown substances reacting with TBA remains to be elucidated.In this work, a derivative spectrophotometric method for MLD determination, based on its reaction with TBA, is described. Owing to the advantageous nature of derivative spectrophotometry for the determination of absorbent ana- lytes, in the presence of several interferent species with overlapped spectra, and for the elimination of the background signal,26 it has been possible to determine MLD selectively in human plasma. Experimental Apparatus A Milton Roy Spectronic 3000 array spectrophotometer provided with Milton Roy software was used for all absorption measurements, storage and analysis of the spectrophotometric data. Differentiation was performed by the simplified least- squares procedure of Savitzky and Golay.27 A thermostatic- ally controlled bath, Selecta Unitronic 320 OR, was used for temperature control.Reagents I, 1,3,3-Tetraethoxypropane (TET) was prepared by dissolv- ing 0.01 g of reagent (Sigma) in 100 ml of water. Malon- appropriate was prepared by generation in situ from TET in acid medium. A 3 x 10-2 mol I-' TBA standard solution was prepared by dissolving 0.45 g of reagent (Sigma) in 100 ml of water and sonicating for 15 min. A 0.01% m/v HMF solution was prepared by dissolving 0.01 g of reagent (Sigma) in 100 ml of water. A 0.01% m/v FUR standard solution was prepared by dissolving 0.01 g of reagent (Sigma) in 100 ml of water. A90 W C m n n 0.40 v) a 0.20 ANALYST, JANUARY 1993, VOL. 118 - - 0.01% m/v GLY solution was prepared by dilution of the appropriate volume of reagent solution (30% in aqueous medium) (Sigma) in 100 ml of water. A 0.01% m/v BIL solution was prepared by dissolving 0.01 g of reagent in 100 ml of water previously made basic.A 0.01% m/v (in each component) ETA-SUC solution was prepared by dissolving 0.01 g of sucrose reagent plus 0.01 g of ethanal reagent (Sigma) in 100 ml of water. Demineralized water was used and all experiments were performed with analytical-reagent grade chemicals. B - 0 e = - - e I I Procedures for Determining MLD General procedure Place an aliquot of TET solution containing an amount of TET equivalent to 1.25-12.5 pg of MLD, 1 ml of 12 rnol I-* hydrochloric acid and 12 ml of 3 x 10-2 moll-' TBA solution in a 25 ml calibrated flask and dilute to the mark with de-ionized water.Heat the samples at 60°C for 60 min in a thermostatically controlled bath and then record the absorp- tion spectra of the samples between 450 and 600 nm and the first-derivative spectra using a AL of 17 nm. Determine the MLD content from the first-derivative spectrum by automatic- ally measuring the peak-to-peak amplitude ,5J3 and using the appropriate calibration graph. Human plasma Direct procedure. Plasma samples were analysed imme- diately after centrifugation. Samples were prepared in 25 ml calibrated flasks to contain 500 pl of human plasma and the General Procedure was followed. Procedure with previous precipitation of the proteins. Acetonitrile (2000 pI) was vigorously mixed with 1000 pl of plasma and centrifuged.The supernatant was filtered through a 0.4 pm filter, an aliquot of 1500 pl was transferred into a 25 ml calibrated flask and the General Procedure followed. Results and Discussion Free MLD is a strongly reactive compound that is unstable in aqueous media and which shows an absorption spectrum in the UV region with a maximum at 244 nm. The MLD reacts with TBA in acidic media and gives rise to a stable and soluble derivatized compound with an absorption maximum located at 532 nm, when the 2 : 1, TBA-MLD adduct is preferentially formed; for low TBA concentrations, a second absorption maximum is located at 395 nm. The absorption spectra for samples containing 0.47 pg ml- 1 of derivatized MLD and free MLD are shown in Fig. 1. An elevated bathochromic shift and a highly sensitive absorption signal are observed for the derivatized MLD.MLD + TBL 0.79 - z e $ n m 4 0.39 - .Q MLD 0 300 400 500 600 700 Wavelengthhm Fig. 1 Absorption spectra of derivatized and free MLD Optimization of the MLD Generation Owing to the instability of aqueous solutions of MLD, standard samples of MLD generated in situ from TET were employed. The optimum chemical and physical conditions for its generation were studied. The rate of MLD generation depends on the acid concentration. The kinetic curves for different hydrochloric acid concentrations at room temperat- ure are shown in Fig. 2. About 2-3 min are sufficient for the development of the reaction using a 1 rnol 1-1 hydrochloric acid concentration. A waiting time of 5 min was selected as sufficient for the MLD generation.A calibration graph was obtained at 244 nm for an MLD concentration of up to 5.95 pg rnl-l, generated in situ from standard samples of TET. The generation was quantitative over the concentration range studied. Study of the TBA-MLD Reaction Optimization of the chemical conditions Several chemical parameters affect the TBA-MLD forma- tion, such as temperature, heating time and acid concentra- tion. Numerous and sometimes drastic chemical conditions have been reportedhJ8.29 for diverse specific applications, but the chemical conditions for the determination of MLD have been studied here, with the aim of selecting the mildest possible chemical conditions for this reaction. A certain amount of MLD can be originated during the reaction with TBA at elevated temperature, from thermal decomposition of fatty peroxide, giving rise to false results in the determination of MLD in real samples.24 The influence of the acidic medium was studied using different hydrochloric acid concentrations for samples con- taining 1.19 pg ml-1 of MLD and 8.4 x 10-3 rnol 1-1 L I I I 0 5.00 10.00 15.00 20.00 Ti m e/m i n Fig.2 rnol I-' HCI; B, 0.5 rnol 1-1 HCl and C, 1.0 rnol 1-1 HCI Kinetic curves for MLD generation in acidic media. A , 0.1 0.80 F 0.60 h AANALYST, JANUARY 1993, VOL. 118 0.400 91 - ?a I 0 -0.100 ' 1 I _. 0 40 80 120 Heating time/min Fig. 4 Influcncc of heating time on the formation of TBA-MLD at A and C, 60 "C and B and D, 80 "C. Absorbance monitored at 531 nm (A and B) and 395 nm (C and D) /I 1.200 0.003 0.008 0.013 Fig.5 Influcncc of TBA concentration on the formation of TBA-MLD. A and D, 0.5 moll-' HCl; B and C. 5.0 moll-1 HCI. 0 = 532 nm; 0 = 395 nm [TBA]/mol I- concentration of TBA; the samples were maintained at 40°C for 30 min. The absorbance at 532 nm was constant between 0.1 and 1.2 moll- 1 hydrochloric acid and decreased for higher concentrations (Fig. 3). The absorbance at 39.5 nm increased slightly when the acid concentration increased. A 0.5 mol 1-1 HCl concentration was selected as optimum for the complete development of the red colour. The influence of the TBA concentration was studied for two different acidity values, 0.5 and 5.0 mol 1-1 HCl. In both media, an increase of the TBA concentration favoured the formation of a species that absorbed at 532 nm; however the absorbance maximum, located at 359 nm, decreased when the TBA concentration increased (Fig.4). A 0.014 mol 1-1 TBA concentration was selected as optimum, because TBA solubil- ity does not allow the use of higher TBA concentrations. Influence of temperature and heating time The effect of temperature on the reaction was examined whilst maintaining a constant heating time of 30 min. The tempera- ture was varied between 20 and 80°C and the samples were cooled to room temperature, before the absorption spectra were recorded. The absorbance at 395 nm was constant over the range of temperatures studied whereas the absorbance at 532 nm increased markedly with temperature up to 6O"C, while the absorbance was constant in the range from 60 to 80°C.The influence of the length of heating time between 20 and 100 min at 60 and 80°C was examined. At 60°C the absorbance was constant for a heating time of between 50 and 100 min, however, at 80 "C the reaction product was unstable for heating times of longer than 50 min (Fig. 5 ) . An optimum heating time of 60 min at 60°C was selected. 1.200 0.800 0 m f! % a a 0.400 0 0.020 1 (b) 0.01 0 I a m .- c .- & O Iz U w .- LL -0.010 -0.020 320 420 520 620 Wavelengthlnm (a) Zero-order and (b) first-derivative absorption spectra of Fig. 6 TBA-MLD in the 0.054.5 pg ml-l concentration rangc Conventional and first-derivative spectral characteristics Under optimum conditions, the derivatized MLD shows an absorption maximum at 531 nm. First-derivative spectra were obtained with different AX values and 17 nm was considered as suitable.The conventional and first-derivative spectra are shown in Fig. 6 for different MLD concentrations. Calibration graphs for MLD concentrations of up to 0.5 pg ml-1 can be established using measurements of the absorbance at 531 nm and the first-derivative peak-to-peak signal 1D521,543. Relative errors of 0.92 and 0.98% ( n = 11, 95% confidence level) were obtained for the conventional and derivative methods, respectively. The equations and the statistical parameters are summarized in Table I. Influence and Elimination of Diverse Interferent Species Study of the interferences Many diverse interferent species are described in the literature for the reaction of TBA with MLD. Some aldehydes produced in hexose degeneration react with TBA, forming pigments that absorb strongly in the visible spectral region and which are, therefore, potential interferents in the determination of MLD with TBA using the conventional method.The chemical behaviour of MLD with TBA in the presence of GLY, HMF, FUR, SUC-ETA mixtures and BIL was investigated under the optimum conditions previously ascer- tained, with the aim of establishing the reaction characteristics of each interferent and hence of avoiding interference. In order to investigate the degree of interference of the above-mentioned species in the reaction of MLD with TBA, the following experiments were carried out. Maintaining a constant concentration of MLD of 0.177 pg ml-1, under optimized chemical conditions and in the presence of the interferent species over a range of concentrations of between 1.8 and 18 pg ml-l (1 : 10 to 1 : 100 MLD: interferent m/m92 ANALYST, JANUARY 1993.VOL. 118 40 -- 30 - 0 - - -10 Table 1 Statistical parameters for thc determination of MLD -'L I I I I I Analytical Linear LOD*/ Method signal Equation regression ng ml-1 Conventional 531 nm A = 2.32[MLD]T + 0.009 0.9990 0.36 Derivative '0521.543 1052i,543 = O.O77[MLD]t + 0.0001 0.9997 1.30 * LOD = limit of dctection. t pg rn1-l. 0.50 r I 0.40 0.30 0.20 0.10 0 0.004 0 .- CI > -0 .- & O 4- r .- L -0.004 I I 375 475 575 675 Wavelengthhm Fig. 7 ( a ) Absorption spectra: A, TBA-MLD (0.17 pg ml-1); B. TBA-FUR (18 pg ml-I): C, TBA-FUR (10 pg ml-I); D, TBA-FUR (1.8 pg mi-1); F, TBA-MLD-FUR (MLD-to-interferent ratio = 1 : 10); and G.TBA-MLD-FUR (MLD-to-interferent ratio = 1 : 100). (b) First-derivative spectra: A, TBA-MLD (0.17 pg ml-1): B, TBA-FUR (18 pg ml-1) and G. TBA-MLD-FUR (MLD-to-inter- ferent ratio = 1 : 100) -0.008 2.000 I I 1.500 0 c: m e 1.000 z n a 0.500 320 420 520 620 Fig. 8 Absorption spectra: A, TRA-MLD (0.17 pg ml-l); B, TBA-HMF (18 pg ml-1); C, SUC-ETA (18 ygml-1); D, TBA-GLY (18 pg ml-1) and E, TBA-BIL (18 pg ml-I) Wavelengthhm Interference leve Is 1 50 1 (a) I s FUR HMF SUC-ETA GLY BIL 2 -Conventional E S Derivative L W 16 14 6 4 2 0 I I I I FUR HMF SUC-ETA GLY BIL = Conventional Derivative Types of interference interferences 71 Vl a : No competitive reaction with TBA b: No reaction with TBA c: Competitive reaction with TBA Fig. 9 Levels and types of interferents in the determination of MLD with TBA.( a ) lnterferent-to-MLD ratio = 100 : 1 and (6) interfcrcnt- to-MLD ratio = 10: 1 ratio), the conventional and first-derivative spectra were obtained against THA solution. The absorption spectrum of the product of the reaction between FUR and TBA is shown in Fig. 7(a), and is compared with the absorption spectra obtained for FUR-TBA in the presence of MLD for a 1 : 100 MLD : FUR ratio. The absorption spectrum obtained for the product of the reaction between TBA and MLD is alsoANALYST, JANUARY 1993, VOL. 118 I 1 I 93 0.600 0.400 0.200 0.600 0.400 0.200 0 1 I I 325 425 525 62 5 325 425 525 625 0.01 0 0.005 o, .- 4- > U .- G o 4- z .- LL -0.005 0.008 o, 0.003 .- 4- > .- G z -0.002 .- LL -0.007 included.The TBA-FUR product exhibits a continuous absorption between 475 and 575 nm and a second absorption maximum is located at about 420 nm. It is well known that derivative spectrophotometry is often used for reducing the effect of background spectral interfer- ences. The first-derivative spectra for the samples mentioned previously are shown in Fig. 7(6). The derivative signal 1Ds21,s43 for FUR-TBA is zero and the values obtained for MLD determination both in the presence and absence of FUR are similar. Behaviour similar to that previously described is also observed in the presence of HMF. The absorption spectrum for the TBA-HMF product shows an absorption maximum located at about 435 nm and a second absorption region in the range between 490 and 620 nm.The absorption spectrum of the GLY-TBA product only shows an absorption maximum in the range between 420 and 490 nm and this product does not absorb at higher wavelengths. However, in the presence of GLY the results obtained for the determination of MLD are less than the expected results; a small relative error is also obtained. In this case, a chemical interference is produced due to the consump- tion of TBA in the reaction with GLY. Similar behaviour is observed for SUC-ETA mixtures, but the relative interfer- ence level is higher. The pigment BIL exhibits continuous absorption through- out the visible spectrum, with absorption maxima located at 630 and 350 nm and it does not react with TBA. Bilverdine produces a spectral interference from overlapping spectra in the conventional TBA method.The absorption spectra for TBA interferents are shown in Fig. 8. Elimination of Interferences and Comparative Study With the aim of eliminating the above-mentioned interfer- ences, the proposed derivative method was applied. The conventional method was also applied and the results obtained were compared with those obtained using the derivative method. The results obtained for two different inter- ferent:MLD ratios and a schematic diagram of the types of interferences found are shown in Fig. 9. In all cases, a very notable decrease of the errors is observed when the derivative method was applied, except for the chemical interference of SUC-ETA. Applications The amount of material that reacts with TBA (MLD plus MLD-like compounds) present in human plasma is described in the literature; 3.8 pmoll-* is obtained using a conventional TBA method, however, MLD remains undetected using a direct HPLC method.12 In this work, the proposed derivative and conventional methods were applied to the determination of MLD in human plasma and a comparative study of proteinized and deproteinized samples was made.Blood was collected from 15 control subjects by venipunc- ture and a pool of the samples was made. After centrifugation, the plasma was immediately divided into two similar aliquots. The first aliquot was deproteinized as described above and the other was not pre-treated. The standard additions method was applied to both samples, proteinized and deproteinized. The absorption and derivative spectra were obtained (Fig.10) and the recovery ratios of known amounts of MLD added to the plasma were calculated for both series by applying both conventional and derivative methods using calibration graphs obtained under the same chemical conditions. The amounts of MLD in the plasma (taking into account that a 1 + 16 dilution of the samples was made) and the interference of the plasma matrix in the determinations were established. The recovery values obtained show that MLD had not been generated during the TBA reaction with plasma samples (Tables 2 and 3). The MLD content obtained using the derivative and94 ANALYST, JANUARY 1993, VOL. 118 Table 2 Results obtained by application of conventional and derivative methods to previously deproteinized human plasma Conventional method" Derivative method? Added/pg ml-1-t Found/pg ml-l$ Recovery(%) Added/pg ml-'-1 Found/pg ml- l$ Recovery(%) 0.00 0.48 - 0.00 0.04 2.95 3.46 101 2.95 3.22 5.90 6.54 103 5.90 6.25 8.85 9.3s 100 8.85 9.03 11.80 12.49 102 3 1.80 12.45 " Slope/pg-I ml cm-1 = 2.31.-1 Slope/Lg-l ml cm - 1 = 0.079. $ A 1 + 49 dilution was made in the original plasma sample prior to measurements. - I07 105 101 105 Table 3 Results obtained by direct application of conventional and derivative methods to human plasma Conventional method" Derivative methodt Added/pgml-l Found& ml-l$ Recovery(%) Added/pgmlkL Found/pg ml- l $ Recovery(%) - 0.00 0.00 - 0.00 0.24 2.95 3.46 110 2.95 2.32 78 5.90 5.80 94 5.90 4.38 74 8.85 9.13 101 8.85 6.71 76 11.80 12.12 101 11 .80 8.77 74 * Slopdpg-1 ml cm-I = 2.31.-1 Slopc/pg-' ml cm - 1 = 0.059. $ A I + 49 dilution was made in the original plasma sample prior to measurements. 0.5 r 0.4 - E 0.3 G = 0.2 0.1 0 Direct Deprotei n ized Sample = Conventional Derivative Cl TBA-test(literature1 E HPLC Fig. 11 using different methods Comparative diagram of the human plasma MLD content conventional methods and the literature data obtained using the TBA test and HPLC are compared in Fig. 11. Conclusions When using the conventional method, appreciable amounts of MLD are measured in human plasma, similar to the amounts described previously;" these results are false because several MLD-like TBA compounds absorb at 532 nm. However, when the proposed derivative method was applied a very low MLD content, less than the limit of detection, was detected in human plasma in accordance with the results described for the HPLC method.12 Adequate recovery values for MLD in deproteinized human plasma samples are obtained and this indicates that the established chemical conditions are adequate to determine MLD in plasma.Smaller recovery values for MLD are obtained owing to interference from precipitation of the sample matrix, when the samples have not been deproteinized. The authors are grateful to the DGICYT of Spain for financial support (Project No. PBSS-0431) and to Dr. F. Henao Davila for helpful discussions. 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 References Miller, D. M., Buettner, G . R., and Aust, S. D., Free Radicals Biol. Med.. 1990, 8, 95. Nair, V., Cooper, C.S . , and Vietti, D. E., LipidA, 1986, 21, 6. Nair. V., Turner, G. A., and Offerman, R . J . , J. Am. Chem. SOC., 1984, 106, 3370. Basu, A . K.. O'Hara, S . M., Valladier, P., Stone, K.. Mols, O., and Marnctt, L. J., Chern. Res. Toxicol., 1988. 1, 53. Steinberg. D., Parthasarathy, S . , Carew. T. E.. Khoo, J. C., and Witztum, J . L., New Engl. J . Med., 1989, 320, 915. Hadley, M., and Draper, H . H . , Free Radicals Biol. Med., 1989, 6, 49. Draper, H. H., McGirr, L. G., and Hadley. M., Lipidy, 1986, 21, 305. Stocks, J., Kemp, M., and Dormandy, T. L., Br. 1. Haemaf., 1971, 20. 95. Kwan, T. W., and Watts, B. M., Anal. Chem., 1963, 35, 733. Bond, A. M.. Deprez, P. P.. Jones, R. D., Wallace, G. G., and Briggs, M. H.. Anal. Chem., 1980. 52, 2211. Sterbauer, H . , and Slater, 7'. F.. IFCS Med. Sci., 1981, 9, 749. Largilliere, C., and Melancon, S. B., Anal. Biochem., 1988, 170, 123. Sinnhuber, R. O., Yu, I . C., and Yu 7'. C., Food RPS., 19.58,23, 620. Sawicki. E.. Stanley. T. W., and Johnson. H., Anal. Chem., 1963, 35, 199. Nair, V.. and Turner, G. A., Lipids, 1984, 19, 804. Yu, L. W.. Latriano. L.. Duncan. S . , Hartwick, R. A.. and Witz, G., Anal. Biochem.. 1086, 156, 326. Kosugi, H., and Kikugawa, K., Lipids, 1986, 21. 537. Kosugi, H.. Kato, T . , and Kikugawa. K., Anal. Biochem., 1987, 165, 456. Kosugi. H., and Kikugawa, K., Free Radicals Biol. Med., 1989, 7. 205. Levillain, P., and Fompeydie, D . , Analusis, 1986. 14, 1. Ottolenghi, A., Arch. Biochem. Biophys.. 1959, 77, 355.ANALYST, JANUARY 1993, VOL. 118 95 22 Asakawa, T., and Matsushita. S., Lipids, 1980, 15, 137. 23 Kosugi. H., Kato, T., and Kikugawa, K . , Lipids, 1988,23,1024. 24 Janero, D. R., Free Radicals Biol. Med., 1990, 9, 515. 25 Stocks, J., Kcmp, M., and Dormandy, T. L., Lancet, 1971, 1, 266. 26 Salinas, F., Espinosa-Mansilla, A . , and Berzas, J. J., Anal. Clzim. Acta, 1990, 233. 289. 27 Savitzky, A., and Golay, M. J . E., Anal. Chern.. 1964,36, 1627. 28 McKnight, R. C., and Hunter, F. E., Biochim. Biophys. Acra, 1965, 98, 640. 29 Ramirez, M. A., and Spillman, D. H . , J. Food Sci., 1987, 52. 500. Paper 210,3876E Received July 21, 1992 Accepted September 10, 1992