Anal. Chem. 1999, 71, 2053-2058
Automated Measurement of Lipid Hydroperoxides in Oil and Fat Samples by Flow Injection Photometry Kang Tian and Purnendu K. Dasgupta*
Department of Chemistry and Biochemistry, Texas Tech University, Lubbock, Texas 79409-1061
The status of oxidation of oil and fat samples is normally judged by their peroxide value (PV), an index that seeks to determine the peroxide content by measuring how much iodine is liberated by the sample from iodide in an acidic medium in a specified time period. At peroxide levels of interest, not only does this approach require considerable analyst skill, the method is inherently flawed because of the considerable differences in the rates at which different peroxides liberate iodine from iodide and potentially also the consumption of nascent iodine by unsaturated sites. We propose here a substantially less biased method based on the oxidation of Fe(II) to Fe(III) by peroxides, followed by the colorimetric detection of the latter as the thiocyanate complex. The system is automated through flow injection photometry. A methanolbutanol mixed solvent is used as the carrier stream. Samples, generally prediluted in the same solvent, are injected into this carrier; streams bearing Fe2+ and SCNare separately introduced, mixed in-line, and merged with the carrier stream. After a reaction time of 30 s, the optical absorbance is detected by a light emitting diode (LED) based photometric detector. The method exhibits a wide dynamic range and good linearity (linear r2, 0.9943 for 0.1-120 mequiv/kg cottonseed oil hydroperoxides) with a good throughput rate (up to 60 samples/h). The method was successfully applied to both vegetable (olive and cottonseed) oils and animal fats (lard and poultry fat).
soluble vitamins and produce toxic byproducts that negatively impact cell turnover in the liver and the immune response of the animal.4,5 The process of lipid peroxidation is believed to be the underlying cause of several diseases such as coronary heart disease caused by hypercholesterolemia,6 cerebral apoplexy,7 and the general process of aging in living animals.8 In the context of foods, the review of the kinetics of lipid oxidation by Labuza9 is dated but still remains the most informative. Lipid peroxidation is a major concern in a variety of products; indeed, the formulation of antioxidants for incorporation in products ranging from potato chips to skin creams is a major area of endeavor.10 Several methods have been developed for determination of lipid hydroperoxides.11 Iodometric methods have been most commonly used for measuring lipid hydroperoxides in oils and fats. The standard method12 involves the reaction of the sample, dissolved in acetic acid/chloroform (3:2 v/v), with aqueous KI for exactly 1 min and then titration with thiosulfate using a starch indicator. Because of concerns with the use of a halogenated organic solvent, a new version of this method utilizes isooctane in lieu of chloroform.13 These methods are time-consuming and relatively insensitive. Because a biphasic system is used, the precise degree of stirring affects results, and careful shielding from light is also necessary.14 Direct UV absorbance measurement of the iodine liberated is substantially more sensitive,15 although there can be potential interference from other UV absorbing species. The fundamental problem with this method, however, is that the kinetics of iodine liberation vary greatly among different hydro-
Oils and fats oxidatively degrade over a period of time. Lipid hydroperoxides are the primary products for this decomposition. They gradually decompose further to secondary products such as aldehydes, alcohols, carboxylic acids, and epoxides. Lipid peroxidation in edible oils and fats has received much attention because it is the primary process of deterioration leading to deleterious changes and rancid odor.1,2 Oxidation of the fat in animal feeds reduces the energy value of the feed and, consequently, increases feed consumption to compensate and, moreover, generates toxic byproducts.3 The peroxides produced by oxidation can destroy other valuable nutrients such as the fat-
(4) Cabel, M. C.; Waldroup, P. W.; Shermer, W. D.; Calabotta, D. F. Poult. Sci. 1998, 67, 1725-1730. (5) Dibner, J. J.; Atwel, C. A.; Kitchell, M. L.; Shermer, W. D.; Ivey, F. J. Animal Feed Sci. Technol. 1996, 62, 1-13. (6) Wang, S. Y.; Bottje, W.; Maynard, P.; Dibner, J. J.; Shermer, W. D. Poult. Sci. 1997, 76, 961-967. (7) Yoshikawa, T.; Yamaguchi, K.; Kondo, M.; Mizakawa, N.; Ohta, T.; Hirakawa, K. Arch. Gerontol. Geriatr. 1982, 1, 209-218. (8) Harman, D. In “Free Radicals in Biology,” Pryor, W. A., Ed.; Academic Press: New York, 1982; Vol. 5, pp 255-275. (9) Labuza, T. P. CRC Crit. Rev. Food Technol. 1971, 2, 355-405. (10) Garewal, H. S., Ed. Antioxidants and Disease Prevention; CRC Press: Boca Raton, FL, 1997. (11) Scheirs, J.; Carlsson, D. J. Technol. Eng. 1995, 34, 97-116. (12) American Oil Chemist’s Society Official Method Cd-8-53, 1996. (13) American Oil Chemist’s Society Official Method Cd-8b-90, 1996. (14) Oishi, M.; Onishi, K.; Nishijima, M.; Nakagomi, K.; Nakazawa, H.; Uchiyama, S.; Suzuki, S. J. AOAC Int. 1992, 75, 507-511. (15) Hicks, M.; Gebicki, J. M. Anal. Biochem. 1979, 99, 249-253. (16) Asakawa, T.; Matsusita, S. J. Am. Oil Chem. Soc. 1978, 55, 619-620. (17) Asakawa, T.; Matsusita, S. Lipids 1980, 15, 965-967.
(1) Kaneda, T.; Ishii, S. J. Biochem. (Tokyo) 1954, 41, 327-335. (2) Moll, C.; Biermann, U.; Grosch, W. J. Agric. Food Chem. 1979, 27, 239243. (3) Steinberg, D.; Parthasarathy, S.; Carew, T. E.; Choo, J. C.; Witztum, J. L. New Engl. J. Med. 1989, 320, 915-924. 10.1021/ac9813181 CCC: $18.00 Published on Web 04/21/1999
© 1999 American Chemical Society
Analytical Chemistry, Vol. 71, No. 10, May 15, 1999 2053
peroxides;15 this has not received much attention in the literature. Various catalysts such as silica gel,16 AlCl3,17 and Fe(II),18 etc. have been used; however, no systematic study has ever been carried out to establish that such approaches are effective for reliably measuring different peroxides in real samples on an equivalent basis. Several highly sensitive enzymatic methods have been reported for determination of hydroperoxides in biological samples.19,20 On the basis of a comparative study of several methods, at least one report21 claims that the enzymatically mediated glutathione-coupled oxidation of NADPH19 is a superior choice for abiologic applications as well. The applicability of the enzymatic methods for determining lipid hydroperoxides in nonaqueous media (as would be necessitated in measuring organic hydroperoxides in oil and fat samples) has never been established. In any case, the sophistication and the cost per assay for these methods will generally preclude their routine use in the edible oil and fat industry. The same criticism applies even more to enzyme-coupled chemiluminescence detection for hydroperoxides, following high performance liquid chromatographic (HPLC) separation.22,23 More recently, Meguro et al. have reported extensively on the determination of hydroperoxides by their reaction with diphenyl1-pyrenylphosphine (DPP) to form the corresponding intensely fluorescent DPP oxide.24 Applications to a variety of samples have been reported, directly25,26 or in the postcolumn reaction format after HPLC separation.27-31 A similar reaction of hydroperoxides occurs with triphenylphosphine (TPP) to form TPP oxide, and lipid hydroperoxides have been measured by FT-IR spectroscopy for this reaction.32 The fluorometric DPP method is indeed sensitive and selective; a version, automated by flow injection (FI), has been reported specifically for use with edible oil and fat samples.33 Nevertheless, this adaptation is stretched because the DPP reaction is slow and therefore not well-suited for FI. The necessary reaction time, even at elevated temperature, is on the order of 5.6-7.6 min, which, in turn, requires 30-50 m reaction coils maintained at 80 °C. To keep dispersion down, narrow bore tubes had to be used, necessitating high-pressure pumps. Overall, the relatively high cost of equipment and the reagent will be a significant drawback for this method to be widely used in the oil and fat industry. A colorimetric FI method has been also reported (18) Løvaas, E. J. Am. Oil Chem. Soc. 1992, 69, 777-783. (19) Heath, R. L.; Tappel, A. L. Anal. Biochem. 1976, 76, 184-191. (20) Marshall, P. J.; Warso, M. A.; Lands, W. E. M. Anal. Biochem. 1985, 145, 192-199. (21) Frew, J. E.; Jones, P.; Scholes, G. Anal. Chim. Acta 1983, 155, 139-150. (22) Yamamoto, Y.; Brodsky, M. H.; Baker, J. C.; Ames, B. N. Anal. Biochem. 1987, 160, 7-13. (23) Miyazawa, T.; Yasuda, K.; Fujimoto, K. Anal. Lett. 1988, 21, 1033-1044. (24) Meguro, H.; Akasaka, K.; Ohrui, H. Methods Enzymol. 1990, 186, 157161. (25) Akasaka, K.; Sasaki, I.; Ohrui, H.; Meguro, H. Biosci. Biotechnol. Biochem. 1992, 56, 605-607. (26) Akasaka, K. Tohoku J. Agric. Res. 1995, 45, 111-119. (27) Akasaka, K.; Ohrui, H.; Meguro, H. Anal. Lett. 1988, 21, 965-975. (28) Akasaka, K.; Ohrui, H.; Meguro, H. J. Chromatogr. 1993, 622, 153-159. (29) Akasaka, K.; Ohrui, H.; Meguro, H. J. Chromatogr. 1993, 628, 31-35. (30) Akasaka, K.; Ohrui, H.; Meguro, H. Biosci. Biotechnol. Biochem. 1994, 58, 396-399. (31) Akasaka, K.; Ohata, A.; Ohrui, H.; Meguro, H. J. Chromatogr., B 1995, 665, 37-43. (32) Dong, J.; Ma, K.; Voort, van de; Frederick, R.; Ismail, A. A. J. AOAC Int. 1997, 80, 345-352. (33) Akasaka, K.; Takamura, T.; Hiroshi, H.; Meguro, H.; Hashimoto, K. Biosci. Biotechnol. Biochem. 1996, 60, 1772-1775.
2054 Analytical Chemistry, Vol. 71, No. 10, May 15, 1999
which relies on the oxidation of a dye substrate by a hydroperoxide.34 Despite the use of an enzyme mimic as a catalyst, this reaction is also slow (requiring 6 min to reach 80% completion for t-BuOOH), and it is very unlikely, therefore, that equivalent responses will be obtained from higher homologues within a reasonable period. The determination of oxidants, including hydroperoxides, by oxidizing Fe(II) to Fe(III) and measuring the latter colorimetrically as the thiocyanato complex, is a very well established reaction.35-39 It is further known that this reaction is compatible with an organic medium like benzene-methanol or chloroform-methanol, that it responds to a broad variety of hydroperoxides, including those of unsaturated fatty acids, and that it is generally more attractive than a competitive method involving the oxidation of Fe(II)-ophenanthroline.40,41 In this paper, we propose a simple, rapid, and sensitive FIA system for the determination of lipid hydroperoxides in oils and fats, where results are not dependent on precise reaction time. It is based on the Fe(III)-thiocyanate complex formation, but careful deoxygenation of reagents is unnecessary, unlike manual versions of this method. It requires no catalysts, heated reactors, highpressure pumps, or excessively long reaction coils. Applications for the analysis of cottonseed oil, olive oil, lard, and poultry fat are demonstrated. EXPERIMENTAL SECTION Cottonseed oil, olive oil and lard were bought from supermarkets. The poultry fat was obtained from a commercial rendering plant. To prepare samples of different peroxide value (PV), the oil/fat sample in a beaker was heated to 100 °C on a hot plate, and air was continuously bubbled through it via a 20-µm porous frit to form a fine stream of bubbles. A substantial amount of sample was periodically withdrawn, and its PV was assayed (according to the standard method,12 except that in later work a potentiometric indication of end point, instead of a visual indicator, was used). The assayed sample was frozen (-15 °C) for later use in the automated assay method. Prior to analysis by the FI method, different amounts of these standardized oil or fat samples (0-1.5 g, of known PV in mequiv/kg) were taken and dissolved in 1:1 (v/v) 1-butanol/methanol and made up to 25 mL. It is to be understood that by the term PV in this paper we merely mean the value determined by the iodometric standard method; the actual peroxide content of the sample might have been different for reasons enumerated later. All chemicals and reagents used were reagent-grade or better and used without further purification. The FI system is shown in Figure 1. Pump P is a Minipuls 2 multichannel peristaltic pump (34) Nakano, T.; Sakida, M.; Miyata, S.; Honda, H. Anal. Sci. 1993, 9, 459465. (35) Lips, A.; Chapman, R. A.; McFarlane, W. D. Oil Soap (Chicago) 1943, 20, 240-243. (36) Chapman, R. A.; MacKay, K. A. J. Am. Oil Chem. Soc. 1949, 26, 360-363. (37) Stine, C. M.; Hartland, H. A.; Caulter, S. T.; Jenness, R. J. Dairy Sci. 1954, 37, 202-208. (38) Koch, R. B.; Stern, B.; Ferrari, C. G. Arch. Biochem. Biophys. 1958, 78, 165-179. (39) Mitsuda, H.; Yasumoto, K.; Iwami, K. J. Jpn. Soc. Food Nutr. 1966, 19, 210214. (40) Petruj, J.; Zehnacker, S.; Sedlar, J.; Marchal, J. Analyst (Cambridge, U.K.) 1986, 111, 671-676. (41) Mihaljevic, B.; Katusin-Razem, B.; Razem, D. Free Radical Biol. Med. 1996, 21, 53-63.
Figure 1. Colorimetric FIA system for the determination of hydroperoxides in oils and fats. See text for details.
(Gilson). The carrier stream CS is 1:1 (v/v) 1-butanol/methanol. Reagents R1 and R2, respectively, consist of 7.7 mM (NH4)2Fe(SO4)2 + 45 mM H2SO4 and 0.13 M NH4SCN, both in methanol. The flow rates in each channel were the same; the system was successfully operated at all flow rates between 0.1 and 1.0 mL/ min (per channel). The presented data pertain to a flow rate of 0.26 mL/min, unless otherwise stated. Solvent flexible pump tubes (Elkay Systems, Shrewsbury, MA) were used for pumping. Valve V is a standard electropneumatically operated 6-port injector (5701P, Rheodyne) equipped with a 20 µL sample loop. The sample S is aspirated by the pump through valve V, any excess going to waste W. The reaction/mixing coil is a 0.7 mm diam × 600 mm poly(tetrafluoroethylene) (PTFE) tube; in conjunction with the connecting tube to the detector and the detector volume, this results in a total residence volume of ∼385 µL and a reaction time of ∼30 s at the cited flow rates. All other tubing used in the flow system is 0.55 mm diam PTFE tubes. Detector D is a homebuilt flow-through photometric detector,42 equipped with a light emitting diode (LED) emitting at 555 nm (HBG5566x, Stanley Electric, Tokyo) with an output signal linearly related to the absorbance; similar detectors are available commercially (Global FIA, Gig Harbor, WA). The data were acquired by a personal computer E (Gateway 2000, P5-75) equipped with a DAS-1601 data acquisition board (Keithley-Metrabyte, Taunton, MA). RESULTS AND DISCUSSION The formation of peroxides, as a function of time, has an induction period and then increases in a sigmoidal fashion because of the autocatalytic nature of peroxide formation. As time continues, the concentration of peroxides reaches a maximum when the rates of peroxide formation and decomposition are equal. This plateau is maintained until the source of fatty acid double bonds has been consumed, at which point the rate of peroxide formation declines. The peroxides already formed will continue to decompose, and the net effect is an eventual decline in peroxide concentration. A typical pattern for the initiation of this sequence is seen in Figure 2. Choice of Detection Wavelength. The colorimetric method is based on the detection of the red complex formed between SCN- and Fe3+. With the very large excess of SCN- present in this system, the absorbance maximum is at ∼510 nm but the (42) Dasgupta, P. K.; Bellamy, H. S.; Liu, H.; Lopez, J. L.; Loree, E. L.; Morris, K.; Petersen, K.; Mir, K. A. Talanta 1993, 40, 53-74.
Figure 2. Temporal profile of peroxide production in cottonseed oil heated at 100 °C and sparged with air.
absorption band is very broad (fwhm, 130 nm). Ultrabright LEDs are available with emission maxima at 495 nm. These would increase the sensitivity by ∼15% relative to the 555-nm LEDs used. However, detection sensitivity was not a problem for any real sample analyzed during this work. In addition, detection at longer wavelengths also increases the immunity against interference from indigenous color in the sample (most typically yellow or brown). Choice of Reaction Medium. Other solvents such as benzene, hexane, ethanol, and ethanol/1-butanol mixtures were studied as potential reaction media both as the carrier stream and as the sample solvent. The 1-butanol/methanol solvent performed better than any of the above. Benzene produced an unstable baseline. Hexane and ethanol yielded much lower sensitivity. Further, hydroperoxides do not show any unusual degradation in 1-butanol/methanol. All the oil and fat samples used in this work, when diluted and stored at room temperature in this solvent, produced identical analytical signals for at least 96 h. Other oil samples that are very susceptible to oxidation, e.g., fish oil, may not be so stable. All samples, especially fats, that are solid at room temperature, may not dissolve totally in the 1-butanol/methanol solvent. The solubility limit for lard, for example, is ∼8 mg/mL. However, the hydroperoxides themselves are polar and readily dissolve in the alcoholic solvent. If the sample is shaken well or heated with the solvent and then allowed to settle, the clear liquid contains the hydroperoxides that have been quantitatively extracted and which can be injected into the FI system without further processing. The extraction is known to be quantitative because a second extraction of the residue contains little or no peroxides. A large amount of fat/oil sample is not needed for analysis because the method is very sensitive. Reagent Concentrations. The two basic steps involved in the acid-catalyzed oxidation of Fe(II) to Fe(III) by a hydroperoxide (43) Swern, D. Organic Peroxides; Wiley-Interscience: New York, 1971; Vol. 2, pp 153-268. (44) Martell, A. E.; Smith, R. M. Critical Stability Constants, Vol. 5; Plenum: New York, 1982; p 401. (45) Sonntag, N. O. V. Bailey’s Industrial Oil and Fat Products, Vol. 1, 4th ed.; Swern, D., Ed.; Wiley: New York, 1979; pp 153-154.
Analytical Chemistry, Vol. 71, No. 10, May 15, 1999
2055
Figure 3. Typical system performance. Lard samples at seven different concentrations are injected in triplicate; the iodometrically measured peroxide values for these samples in µeq/L are shown atop each set of peaks.
are believed to
be:40,41,43
ROOH + Fe2+ + H+ f RO• + Fe3+ + H2O
(1)
RO• + Fe2+ + H+ f ROH + Fe3+
(2)
These reactions would be facilitated by an increased concentration of Fe2+. Mohr’s salt, (NH4)2Fe(SO4)2, is the most stable among simple ionic Fe(II) salts and is therefore the preferred reagent. However, (NH4)2(SO4) begins to precipitate in our solvent system around a concentration of ∼8 mM, especially as a result of the common ion effect from H2SO4. This can cause spurious noise in the detector and blockage of system conduits when the system is shut down. The recommended concentration of 7.7 mM provides some margin of safety. Fe3+ forms a series of thiocyanato complexes Fe(SCN)n(3 - n)+ up to n ) 4.44 The absorption intensity increases, and the absorption maxima shift to longer wavelengths as n increases. To ensure maximum signal intensity and absorption at the longest possible wavelength, we used a NH4SCN concentration of 0.13 M; this concentration is sufficient to ensure maximum absorption.40 System Performance. Typical system output is shown in Figure 3 for Lard samples. Similar output was observed for all the other samples. Cottonseed-oil samples were most extensively studied, and precision of measurements for replicate samples (n ) 7) was studied for PV ranging from 0.03 to 2.5 mequiv/L and ranged from 0.67 to 1.4% in relative standard deviation (RSD). On the basis of the baseline noise, the limit of detection (LOD, S/N ) 3) was computed to be 2 µeq/L. This means if 1 g of an unknown oil sample is dissolved in 25 mL of 1-butanol/methanol solvent, the LOD of hydroperoxides in the original sample will be 50 µeq/kg. To put this value in perspective, the maximum acceptable hydroperoxide concentration in an edible oil/fat sample is generally considered to be 20 mequiv/kg.45 In absolute amounts, the detection limit is 40 peq of hydroperoxide in a complex sample; 2056 Analytical Chemistry, Vol. 71, No. 10, May 15, 1999
Figure 4. Signal is not affected by increasing reaction time in the system. It appears to be complete.
this is very respectable for a colorimetric detection system. The response was also linear with respect to the PV (r2 ) 0.9942). The precision for poultry-fat samples (original samples were dark brown in color) was worse than that for cottonseed oil, being 2.8% at 0.2 mequiv/L and 1.6% at 0.7 mequiv/L (these concentrations refer to the diluted sample actually injected into the system). For highly colored samples, a blank measurement is recommended, with the NH4SCN reagent being substituted with pure methanol. However, even with this highly colored sample, the blank response was quite small (
