Anal Bioanal Chem (2002) 372 : 644–648 DOI 10.1007/s00216-001-1228-0

S P E C I A L I S S U E PA P E R

Erhard Schulte

Determination of higher carbonyl compounds in used frying fats by HPLC of DNPH derivatives

Received: 18 September 2001 / Revised: 22 November 2001 / Accepted: 26 November 2001 / Published online: 2 February 2002 © Springer-Verlag 2002

Abstract In current analytical practice, simple methods are required for assessing the quality of a deep-frying fat after some time of use. Therefore, a new procedure was developed for the fast and selective determination of higher carbonyl compounds, i.e., those triglycerides resulting from the oxidation and/or oxidative cleavage of double bonds in unsaturated fatty acids. For analysis, a fat sample is dissolved in a 1-butanol/toluene mixture containing 6-undecanone as an internal standard. Aldehydes and ketones present are allowed to react with 2,4-dinitrophenylhydrazine (DNPH) in acidic solution for at least 1 h at room temperature. The mixture is injected directly onto a reversed phase HPLC column that is eluted with a very steep gradient between methanol and tert-butylmethyl ether (TBME) and is fitted with a UV detector set at 370 nm. In this way, the DNPH derivatives of all higher carbonyl compounds are eluted as one single peak. The result is calculated as milliequivalents of carbonyls per kg of fat (meq kg–1). The method takes a minimum of time and reagents and only requires the usual equipment. Keywords Carbonyls · Carbonyl compounds · Deep-frying fat · HPLC · DNPH derivatives

Introduction Deep-frying at 130–200 °C in the presence of water and air destroys part of the triglycerides by oxidative and hydrolytic reactions [1, 2, 3, 4]. Hydrolysis is of minor significance, whereas oxidation seems to be much more important [1]. To assess the quality of a used fat, several test methods may be applied, such as the unpleasant and subjective sensory test, the smoke point, the contents of free and oxidized fatty acids, several commercially available

E. Schulte (✉) Institute of Food Chemistry. University of Münster, Corrensstr. 45, 48149 Münster, Germany e-mail: [email protected]

tests (e.g., Oxifrit for some oxidized fatty acids [5] or Fritest for some carbonyl compounds [6]) and density, viscosity, and refractive index, etc. At present and in most cases, the determinations of polar compounds by column chromatography [7, 8, 9, 10] or HPLC [11], and of oligomeric triglycerides by gel permeation chromatography (GPC) [12] are performed in order to complement the sensory findings. When non-polar and polar compounds are determined on a silica gel column, the polar fraction includes all products from hydrolysis, oxidation, and oligomerization. On the other hand, GPC separates monomeric triglycerides from di- and oligomerized molecules which amount to roughly half of the polar silica gel fraction. Many years ago, several methods were proposed for derivatizing carbonyls with 2,4-dinitrophenylhydrazine and measuring the color intensity of the hydrazones [13, 14, 15] as a means for estimating rancidity. However, they all suffer from several drawbacks caused by the reagent and they do not differentiate between short-chain aldehydes and the higher carbonyl compounds. Therefore, we intended to develop an economical method which should be suitable for routine use in order to determine the higher carbonyl compounds as a measure of quality of used deep-frying fats. In these fats, short-chains are already removed by distillation at the high temperatures of the deep-frying, especially in the presence of the water. However, there is no practicable and reliable method available for determining higher carbonyl compounds, which above all, are typical of used frying fats and are suspected to reduce digestibility [1]. Most of these are the so-called core aldehydes [16, 17, 18] (Fig. 1) and they originate from the oxidation of an unsaturated fatty acid of a triglyceride and the subsequent cleavage of the double bond. This results in a volatile short-chain aldehyde and at least one fatty acid containing a terminal aldehyde group in the remaining triglyceride. For determining such core aldehydes and similar carbonyl compounds, it was promising to convert them to the dinitrophenylhydrazones, since the derivatization with 2,4-dinitrophenylhydrazine (DNPH) is also frequently ap-

645 Fig. 1 Chemical structure of a typical core aldehyde, C46H86O7

HPLC conditions Injection volume: 10 µL. Eluent: 0.75 mLmin–1, gradient elution 3min (low pressure mixing), methanol (3 min) −→ tert-butylmethyl 0min ether (3 min) −→ methanol, at least 6 min before the next injection. Detector: UV-detector or DAD at 370 nm.Calculation of results CA =

plied for analyzing carbonyls in other areas [19, 20, 21, 22]. Our particular aim was to allow all of the fat sample to react with DNPH and then to separate the derivatives of the target analytes from the derivatives of short-chain aldehydes, the excess of reagent, and other compounds by using HPLC with UV detection. In view of the highly complex mixture of carbonyls to be expected and the problem of separating all the individual components, we instead tried to find suitable HPLC conditions that would result in only one peak, easily integrated for the sum of all analytes.

Experimental Samples Samples of used frying fats showing all levels of quality reduction were collected from restaurant kitchens by food inspectors. Chemicals All chemicals were of analytical grade. Solvents: 1-butanol, toluene, diethyleneglycol dimethyl ether, methanol, tert-butylmethyl ether. Reagents: hydrochloric acid (37%), 2,4-dinitrophenylhydrazine (DNPH, wet). Internal standard solution: ca. 10 mg 6-undecanone (dipentylketone), exactly weighed, were dissolved in exactly 100 mL 1-butanol/toluene (2:1, v/v). Derivatization reagent: 600 mg DNPH (wet) were dissolved in 8 mL diethyleneglycol dimethyl ether and 1.6 mL hydrochloric acid were added. Accessories Autosampler vials (1.5 mL) with screw caps and PTFE-lined septa. Instrumentation HPLC equipment (Hewlett-Packard/Agilent, Waldbronn, Germany): Autosampler. Gradient pump (low pressure mixing), DAD. Column (Merck, Darmstadt, Germany): LiChrospher RP-18e, 5 µm, 250×4 mm with RP-18e precolumn 4×4 mm. Procedure Exactly weigh approx. 100 mg of fat sample into an autosampler vial (firstly melt the solid fat at about 50 °C) and dissolve it in 1 mL of internal standard solution. To the solution add 250 µL of derivatization reagent and allow the mixture to stand at room temperature for at least 1 h. Inject a suitable aliquot into the high-performance liquid chromatograph.

AA · E∗ A A · E ∗ · 1000000 = · 5872 AI S · F W · E AI S · E

CA, concentration of higher carbonyls in meq kg–1 of fat; AA, area of analyte; AIS, area of internal standard; E*, with 1 mL of internal standard added (6-undecanone) in mg (ca. 0.1 mg); FW, molecular weight of internal standard (170.3 for undecanone); E , sample weight in mg (ca. 100 mg)

Results and discussion Derivatization Several reaction mixtures, times, and temperatures were tested to find out the optimum conditions for derivatizing those carbonyls which may occur in frying fats. Our main concern was to ensure room temperature was maintained during derivatization, thus avoiding the formation of additional artifacts, to obtain high yields of derivatives even in the presence of an excess of other lipids, and to obtain a clear solution that remained stable for several hours at least. DNPH is commercially available with approx. 30% water content and is poorly soluble in water and most organic solvents. This is insignificant in the presence of an acid that is in any case required to catalyze the derivatization reaction. Best results were obtained with a concentrated (approx. 0.2 M) solution of moist DNPH in diethyleneglycol dimethyl ether, acidified with approx. 2 M of hydrochloric acid. As a basic solvent for the fat, 1-butanol proved to be the most suitable since it has a strong solving power for lipids and is compatible with our reversed phase HPLC conditions. To enhance the solubility of solid fats, two parts of 1-butanol were mixed with one part of toluene. In this way, a clear solution was obtained in all cases, both before and after the addition of the derivatization reagent. When this was injected into the HPLC loop, a good separation was achieved despite the very different polarity of the eluent and the presence of a large excess of fat. All solvents and reagents were commercially available and used without further purification to improve the acceptance of the method for routine work. Derivatizing at ambient temperature for at least 1 h resulted in chromatograms in which the peaks were stable for up to approx. 12 h. After this time, the peak of the internal standard remained stable, but the response for the carbonyls in question decreased, possibly due to degradation of the derivatives. The same effect was observed when the mixture was heated to 70 °C for 15 min and consequently, room temperature was the most suitable.

646

HPLC

Calculation of the results

For elution, a gradient from methanol to tert-butylmethyl ether was used. A more gentle gradient [e.g., methanol/wa15 min 15 min ter (2:1, v/v) −→ methanol −→ methanol/tert-butylmethyl ether (1:3, v/v)] resulted in very complex chromatograms with additional broad patterns of solvent and reagent impurities as well as the short-chain aldehyde derivatives. For the higher carbonyls in question, a multitude of peaks was obtained, derived mainly from the core aldehydes, but also from countless other carbonyls, their homologues, and isomers. Since there was no chance to identify individual components, we decided to follow an opposite approach: When using a steep gradient with a very fast change of solvent ratio, all higher carbonyl derivatives were eluted at nearly the same retention time. In this way one single peak was obtained in a short time for all analytes in question (Fig. 2). This peak can be easily integrated and quantitatively evaluated when compared with an internal standard that compensates for varying reaction conditions. For this, 6-undecanone was selected; this is eluted just before the higher carbonyls group and is much more stable than an aldehyde. By using a shorter column it should be possible to reduce analysis time and solvent consumption a little, without loosing quality of quantitative results.

In the literature, the results of other carbonyl determination methods are given as millimoles of carbonyls per kg of fat. These are totally identical with the milliequivalents calculated here, however the latter definition seems more correct, because it can not be excluded that more than one carbonyl group are present in one molecule. The first relation would be correct, if it was given as millimoles DNPH consumption per kg of fat. Because of the complex composition of higher carbonyls, a certain compound for calibration is not realistic. A response factor of 1 is assumed for calculation according to convention. It is possible to calculate a “hypothetical” content of higher carbonyls in mg kg–1 by multiplying the mg kg–1 value by the molecular weight of a core aldehyde (Fig. 1), i.e., C46H86O7 with a FW of 751. From the meq kg–1 value, the percentage (g 100 g–1) is obtained = 5.26% . The whole by dividing by 10,000, e.g., 70meq·751 10000 method is specific for carbonyls. It is independent of color in the fats, since they give virtually no blank if dark fats are analyzed by HPLC without addition of DNPH reagent.

Fig. 2 HPLC separation of a used frying fat spiked with 6-undecanone (IS) as internal standard after reaction with DNPH (A) and corresponding blank without fat (B). HC, higher carbonyls. RP-18e column, 250×4 mm. Eluent: 3min methanol (3 min) −→ tert0min butylmethyl ether (3 min) −→ methanol, minimum 6 min before next injection, 0.75 mL min–1. UV-detection at 370 nm

Validation of the method The “true” values are unknown, since no other independent method exists that is able to determine the same ana-

647

lytes objectively. The older photometric methods [13, 14, 15] using different DNPH reagents and calculation methods do not give comparable results. The oxime-titration is not quite sensitive and reliable enough. So only the coefficient of variation can be determined and this amounted to 1.2% for 10 analyses. Advantages of the method: • • • • • • •

little manual work required results obtained rapidly low amounts of chemicals required equipment present in most laboratories automatic separation and calculation possible selective to carbonyl compounds class of artifacts is determined (relevant for quality)

Results The method was applied to 71 samples of used frying fats showing very different quality. The results are presented in two diagrams (Figs. 3 and 4) where the content of higher carbonyls (in meq kg–1) found for each sample is plotted against the content of polar compounds and oligomers, respectively. The content of polar compounds is today frequently used as a parameter for assessing fat quality and was determined by a gravimetric micromethod using silica gel column chromatography [9]. As can be seen, there is only a low correlation, due to the fact that the fraction of polar compounds contains all polar products including free fatty acids, mono-, and di-glycerides. The ratio of oxidized to hydrolyzed compounds depends on the frying conditions, the composition of the frying fat, and to a minor extent, on the lipids which will migrate from the fried

Fig. 4 Diagram of the contents of higher carbonyls in meq kg–1 fat in 71 samples of deep-frying fats analyzed with this method, plotted against the contents of oligomeric triglycerides per 100 g fat, determined by GPC [12]

goods into the frying fat. Consequently, both parameters are independent to a certain degree, but all higher carbonyls were found in the polar fraction. However, a method for differentiating hydrolysis and oxidation products in used frying fat does not exist to verify this fact quantitatively. The determination of oligomeric triglycerides by high pressure gel permeation liquid chromatography (GPC) is an economical method for routine use [12]. Oligomers are formed in frying fats mainly by oxidation processes as well as the carbonyls do. A much better correlation therefore results (Fig. 4). The non-linear curve fit crosses zero. In contrast to this, the curve in Fig. 3 intersects the x-axis at about 5.6, since new frying fats already contain a certain proportion of polar compounds. The contents found for higher carbonyls ranged from 2–200 meq kg–1. The limit of acceptance is 24 g 100 g–1 fat for the polar compounds and 13 g 100 g–1 for oligomers, which will roughly correspond to 70 meq kg–1 for both. Assuming that the higher carbonyls consist only of one core aldehyde (Fig. 1) with a molar mass of 751 g mol–1, this would result in a content of 5.26 g 100 g–1. Such a high concentration is quite surprising and has not been reported before for a frying fat. Acknowledgements I thank Ms M.-E. Lenczyk, Ms M. Gregoritza, and Mr K.-H. Scheele for the experimental work and Dr C. Gertz for providing the fat samples and for valuable discussions.

References Fig. 3 Diagram of the contents of higher carbonyls in meq kg–1 fat in 71 samples of deep-frying fats analyzed with this method, plotted against the contents of g polar compounds per 100 g fat, determined by microcolumn chromatography [9]

1. Sánchez-Muniz FJ, Bastida S, Gonzáles-Muñoz MJ (1999) Lipids 34:1187–1192 2. Gray JI (1978) J Am Oil Chem Soc 55:539–546 3. Peers KE, Swoboda PAT (1979) J Sci Food Agric 30:876–880 4. Márquez-Ruiz G, Dobarganes C (1996) J Sci Food Agric 70: 120–126

648 5. Meyer H (1979) Fette Seifen Anstrichm 81:524–532 6. Pardun H, Blass J, Kroll E (1974) Fette Seifen Anstrichm 76:151–158 7. Association of Official Analytical Chemists (AOAC) (1998) Official methods of analysis. AOAC International, Gaithersburg/USA, Method 982.27, chap 41.1.34 8. Bundesinstitut für den gesundheitlichen Verbraucherschutz und Veterinärmedizin (BgVV) (1997) Amtliche Sammlung von Untersuchngsverfahren. Beuth Verlag, Berlin, Wien, Zürich, Methode L 13.07.12–1 9. Schulte E (2000) Eur J Lipid Sci Technol 102:574–579 10. Gertz C, Klostermann S (2000) Eur J Lipid Sci Technol 102: 329–336 11. Kaufmann A, Ryser B, Suter B (2001) Eur Food Res Technol 213:372–376 12. Association of Official Analytical Chemists (AOAC) (1998) Official methods of analysis. AOAC International, Gaithersburg/USA, Method 993.25, chap 41.1.67

13. Berry NW, McKerrigan AA (1958) J Sci Food Agric 9:693– 701 14. Pardun H (1976) Analyse der Nahrungsfette. Verlag Parey, Berlin, Hamburg, pp 229–231 15. Chiba T, Takazawa M, Fujimoto K (1989) J Am Oil Chem Soc 66:1588–1592 16. Kuksis A (1990) Inform 1:1055–1060 17. Kuksis A (2000) Inform 11:746–752 18. Kamal-Eldin A, Márquez-Ruiz G, Dobarganes C, Appelqvist LA (1997) J Chromatogr A 776:245–254 19. Kim SS, Gallaher DD, Csallany AS (1999) Lipids 34:489–496 20. Grosjean E, Green PG, Grosjean D (1999) Anal Chem 71: 1851–1861 21. Vogel M, Büldt A, Karst U (2000) Fresenius J Anal Chem 366: 781–791 22. Bartos J, Pesez M (1979) Pure Appl Chem 51:1805–1814