The Journal of Supercritical Fluids

Transcription

1 J. of Supercritical Fluids 49 (2009) Contents lists available at ScienceDirect The Journal of Supercritical Fluids journal homepage: Optimization of enzymatic synthesis of cocoa butter analog from camel hump fat in supercritical carbon dioxide by response surface method (RSM) Hajar Shekarchizadeh a, Mahdi Kadivar a,, Hasan S. Ghaziaskar a, Marzieh Rezayat b a Department of Food Science and Technology, Isfahan University of Technology, Isfahan 84156, Iran b Department of Chemistry, Isfahan University of Technology, Isfahan 84156, Iran article info abstract Article history: Received 2 January 2009 Received in revised form 6 March 2009 Accepted 13 March 2009 Keywords: Enzymatic interesterification Camel hump fat Lipozyme TL IM Supercritical carbon dioxide (SC-CO 2) RSM Supercritical carbon dioxide (SC-CO 2 ) has been studied as a medium for esterification of camel hump fat and tristearin in producing cocoa butter analog using Immobilized Thermomyces lanuginosus lipase (Lipozyme TL IM) as a biocatalyst. Process conditions (pressure, temperature, tristearin/camel hump fat ratio, water content, and incubation time) were optimized by conducting experiments at five different levels using the response surface method (RSM). A second-order polynomial response surface equation was developed indicating the effect of variables on cocoa butter analog yield. Contour maps generated using the response surface equation showed that all the experimental variables significantly affected the yield. The pressure, 10 MPa; temperature, 40 C; SSS/CHF ratio, 1:1; water content, 10% (w/w); and incubation time, 3 h were found to be the optimum conditions to achieve the maximum yield of cocoa butter analog Elsevier B.V. All rights reserved. 1. Introduction Fat constitutes one-third of the chocolate content and its properties affect certain parameters of the chocolate manufacturing process, especially at the tempering and cooling stages [1]. Melting fast and completely in mouth while hard and brittle at room temperature, cocoa butter is the fat phase in chocolate and is characterized by sharp melting point, desirable physicochemical properties, and fatty acid components [2]. It is composed predominantly (>70%) of symmetrical triacylglycerols, 1,3-dipalmitoyl- 2-oleoylglycerol (POP), 1(3)-palmitoyl-3(1)-stearoyl-2-monoolein (POS), and 1,3-distearoyl-2-oleoylglycerol (SOS), with oleic acid in the sn-2 position [3]. The typical fatty acid composition of cocoa butter in mole percentage is: C16:0 24.4%, C18:0 33.6%, C18:1 37.0%, C18:2 3.4%, and others 1.6%. Cocoa butter melts at temperatures between 32 and 35 C and is able to re-crystallize during processing to a stable crystal mode [4]. World production of cocoa butter was about 675,000 metric tons in Numerous factors, however, including the degree of uncertainty in supply, variability in quality, and price premium compared to other fats have driven the search for alternatives [5]. Cocoa butter replacers (CBRs) are those fats which replace cocoa butter partially or wholly in chocolate products. CBRs are further classified as (i) cocoa-butter-equivalents (CBE), which behave like cocoa butter and are able to blend in cocoa butter in any proportion without altering the physicochemi- Corresponding author. Tel.: ; fax: address: mak120@mail.usask.ca (M. Kadivar). cal characteristics of cocoa butter; and (ii) cocoa-butter-substitute (CBS), which are fats that can be blended with cocoa butter to a limited extent, without significantly altering its physicochemical characteristics [6]. Procedures of chemical interesterification for analogs have been developed for years from cheap fats and oils and are found to be practical to industries [7]. Recently, preparation of CBE through 1, 3-specific lipase-catalyzed interesterification has received much attention because lipases offer certain advantages over other chemical catalysts [8]. While chemical catalysts will randomize all of the fatty acids in a triacylglycerol mixture, 1, 3-specific lipase can incorporate fatty acids into the sn-1, 3-positions without changing the fatty acid residues in the sn-2-position [9]. Other advantages include the mild conditions under which the reactions take place, thereby requiring minimal energy inputs, reduced levels of by-products generated during the reaction, and more efficient conversion of labile substrates. In order to improve the reuse and stability of lipase, immobilization of the lipase is necessary [10]. However, CBE prepared through the lipase-catalyzed interesterification of triacylglycerols in organic solvent systems are suffering from some major drawbacks such as final solvent removal, which is usually costly and lengthy. Supercritical fluids, with less mass transfer resistance than the conventional liquid solvents, have been used as solvents for enzymatic reactions [11]. In addition, the solubility of most organics in supercritical fluids is higher than that in gaseous phase and is comparable with liquid solvents. Among supercritical fluids, SC-CO 2 is especially advantageous: low viscosity, high diffusivity and low surface tension, which allow it to penetrate easily throughout macro- and microporous materials /$ see front matter 2009 Elsevier B.V. All rights reserved. doi: /j.supflu

2 210 H. Shekarchizadeh et al. / J. of Supercritical Fluids 49 (2009) [12]. Also, it is a clean technology due to its non-toxic and environmental friendly nature and requires mild operating conditions [13]. Further, its critical point of 7.4 MPa and 31 C is sufficiently low for processing of heat-liable materials. Most importantly, it would be readily separable from the reaction medium simply by post-reactional depressurization. The novelty of the process lies in this easy post-reactional separation of the products to high purity without extensive separation processes, thus offering industrial advantages [14]. The camel stores its energy reserves in the form of fat in different depots in their body of which the hump comprises a considerable amount of the adult body weight; therefore, camels can survive long periods without feed. Hump contains mixtures of fatty acids and most of these are C16:0, C18:0 and C18:1 [15]. Inthe north of Africa, the Middle East, and the south of America, camels are slaughtered for their meat and the fat is mostly consumed directly. Response surface methodology (RSM) is a mathematical tool, which can help in arriving at optimum conditions for a reaction with minimum number of experiments to obtain statistically acceptable results. RSM enables evaluation of the effects of multiple parameters, alone or in combination, on response variables and also predicts their behavior under given sets of conditions [16]. To date, no research has been reported on using camel hump fat (CHF) to produce cocoa butter analog. The objective of this study was to optimize the reaction time, temperature, pressure, tristearin/camel hump fat ratio, and water content by RSM for maximum yield of cocoa butter analog produced by enzymatic interesterification of camel hump fat and tristearin using Lipozyme TL IM under SC-CO Materials and methods 2.1. Materials Immobilized Thermomyces lanuginosus lipase (Lipozyme TL IM) was the gift provided by Novozymes (Tehran, Iran). Fatty acids and tristearin were purchased from Sigma Chemical Co. Tetrahydrofuran and isopropanol used for HPLC analysis were spectral grade and obtained from Merck. Other chemicals and solvents were all analytical grades also purchased from Merck. Cocoa butter was purchased from Cocoa Tabriz Co. (Tabriz, Iran). The purity of CO 2 was more than mass% and was purchased from Zamzam Co. (Isfahan, Iran) Enzymatic interesterification The experimental apparatus with SC-CO 2 as the reaction medium is schematically shown in Fig. 1. Cocoa butter analog synthesis was carried out in a SC-CO 2 batch reactor, a 40-mL high pressure vessel into which camel hump fat and tristearin (SSS) were blended at various weight percent ratios. Then, immobilized lipase Fig. 1. Schematic diagram of supercritical carbon dioxide system for enzymatic reaction. PI, pressure indicator. Table 1 Experimental range and values of the independent process variables in the design for production of cocoa butter analog. Independent variables Symbols Range and levels Reaction pressure (MPa) X Reaction temperature ( C) X Reaction time (h) X Water (%) X SSS/CHF ratio X :1 0.25:1 0.5:1 1:1 2:1 (10% by total weight of substrate) and water at various weight percents of substrate were added. Reactions were carried out under varied conditions of reaction pressure, temperature, and incubation time. After reaction to the desired time, depressurization and elution by n-hexane was conducted. The reaction mixture was filtered immediately through a 0.45 m filter to remove the enzyme; hexane was removed by rotary vacuum evaporator RP-HPLC analysis HPLC was carried out with a Hewlett-Packard model 1090 high-performance liquid chromatographer equipped with an auto sampler, quaternary pump module, and a UV/vis scanning (diode array) detector. The TAG mixtures were separated by a non-aqueous spherical RP-C18 column (150 mm 3.9 mm 5 m). Separations were carried out with acetonitrile/tetrahydrofuran (75:25, v/v as solvent A) and water (as solvent B) as eluents at a flow rate of 1 ml/min and the following solvent gradient profile: initial condition 60:40 (volume percent A/B), to 100% A over 10 min, held for 15 min, and then returned to original conditions over 5 min (the total run time was 30 min). The UV spectra of TAG were determined at 220 nm. The oven temperature was set at 40 C. The samples were prepared at 25 mg/ml in tetrahydrofuran and 20 L of the solution was injected. The peak identification was done by comparing the retention time with those of TAG standards Yield of cocoa butter analog Yield of cocoa butter analog prepared under each set of conditions was calculated from the following equation: yield of cocoa butter analog (%) = mole of cocoa butter analog 100% mole of substrate (CHF and SSS) 2.5. Experimental design A statistical method, central composite designs (CCD), was adopted to optimize the reaction conditions for maximum production of cocoa butter analog from CHF enzymatically. Pressure of the reaction X 1 (MPa); temperature of the reaction X 2 ( C); incubation time X 3 (h); water content, X 4 as percentage of the reactants, and X 5 mole ratio of SSS to mole of CHF were used as the variables to maximize the response (yield of cocoa butter analog). The experimental design included 32 experiments of five variables at five levels ( 2, 1, 0, +1, +2). Table 1 gives the range of variables employed. The actual set of experiments performed (experimental runs 1 32) and the yield of cocoa butter analog obtained are shown in Table 2. A second-order polynomial equation was developed to study the effects of variables on the yield. The equation indicates the effect of variables in terms of linear, quadratic, and cross-product terms: Y = ˇ0 + ˇiX i + ˇii X 2 i + ˇij X i X j

3 H. Shekarchizadeh et al. / J. of Supercritical Fluids 49 (2009) Table 2 Responses of dependent variables for production of cocoa butter analog to independent variables. Exp. no. Variable levels Yield (%) X 1 X 2 X 3 X 4 X 5 Experimental Predicted below: Fig. 2. Normal probability of the cocoa butter analog yield at residuals. Y = X X X X X 1X X 3 X X 4 X 5 The analysis of variance and error for the response surface model are given in Table 4. These results show that the model is significant at P < 0.1 and the linear and square terms are significant at P <0.1. The interaction terms are not significant. The lack-of-fit or adequacy test is significant at P < 0.1 showing the adequacy of the quadratic model selected. The plots of the quadratic model with three variables kept at constant level and the other two varying with the experimental ranges are shown in Figs Effect of pressure where Y is the yield of cocoa butter analog (%), X i and X j are the levels of variables, ˇ0 the constant term, ˇi the coefficient of the linear terms, ˇii the coefficient of the quadratic terms, and ˇij the coefficient of the cross-product terms. All the experimental data were statistically analyzed by Minitab software, version The graphical representations of the above equation in the form of surface plots were used to describe the individual and cumulative effects of the test variables on the response. 3. Results and discussion 3.1. Analysis of the model The cocoa butter analog yields obtained from all the experiments are listed in Table 2 according to RSM design. The predicted response values slightly deviated from experimental data. The normal probability at residuals (Fig. 2) indicated no abnormality in the methodology adopted (R 2 = 0.979). Multiple regression coefficients, obtained by employing a least squares technique to predict a second-order polynomial model for the cocoa butter analog yield, are summarized in Table 3. The significance of each coefficient was determined by F-value and P-value. Corresponding P-values suggest that, among the test variables used in this study, x 1 (pressure), (x 1 ) 2 (pressure pressure), (x 2 ) 2 (temperature temperature), (x 4 ) 2 (water percent water percent), x 1 x 5 (pressure SSS/CHF ratio), x 3 x 4 (time water percent), and x 4 x 5 (water percent SSS/CHF ratio) are significant model terms with P-values of less than 0.1. Other terms are insignificant. The coefficients of independent variables determined for the second-order polynomial model for the cocoa butter analog yield is given The contour plots (Fig. 3) indicate that at a pressure of around 100 MPa, the maximum cocoa butter analog yield was obtained with the other variables at constant levels. The intrinsic effect of pressure on enzymes is an important factor. It has been reported that change in pressure can both directly and indirectly affect enzyme activity, specificity, and stability. Table 3 Regression coefficients of predicated second-order polynomial model for the response variable. Variables Degree of freedom Coefficients t-value P-value Constant a A a B NS C NS D NS E NS A A a B B b C C NS D D b E E NS A B NS A C NS A D NS A E a B C NS B D NS B E NS C D b C E NS D E b NS: not significant at P >0.1. a Significant at P < b Significant at P <0.1.

4 212 H. Shekarchizadeh et al. / J. of Supercritical Fluids 49 (2009) Table 4 Analysis of the variance (ANOVA) for the fit of the experimental data to response surface model. Source Degree of freedom Adjusted sum of squares Adjusted mean squares F-value P-value Regression b Linear b Square b Interaction NS Residual error Lack-of-fit a Pure error Total 31 NS: not significant at P <0.1. a Significant at P < b Significant at P <0.1 Pressure could induce modification of both structure and function of enzyme particles leading to a change in the enzyme activity even though the role of SC-CO 2 molecules during the catalytic step is not yet well understood. Ikushima et al. [17] reported on the possible effect of CO 2 molecules on lipase activity. They studied secondary structures of an enzyme in SC-CO 2 and interactions operating between carbon dioxide and enzyme molecules. It was found that the SC-CO 2 medium in the near critical region should be a trigger to the activation of the enzyme by causing the movement of surface groups and creating an active site, producing stereoselective machinery. In a corresponding manner, other authors stated that enantioselectivity increased with pressure, the maximum value being reached at 10.3 MPa. Afterwards the enantioselectivity decreased with pressure. They thought that the enantioselectivity decreased at pressures above 10.3 MPa indicating a direct inactivation of the enzyme by formation of carbamates [5]. It is proposed that the large change in density could significantly change the interaction of CO 2 and enzyme, causing formation of carbamates from carbon dioxide and free amine groups on the surface of the enzyme. Formation of carbamates leads to enzyme inactivation [18]. Fig. 3. The contour plots: (a) mutual effect of pressure and time, (b) mutual effect of pressure and water, (c) mutual effect of pressure and SSS/CHF ratio and (d) mutual effect of pressure and temperature on the yield of cocoa butter analog.

5 H. Shekarchizadeh et al. / J. of Supercritical Fluids 49 (2009) Fig. 4. The contour plots: (a) mutual effect of temperature and water, (b) mutual effect of temperature and SSS/CHF ratio and (c) mutual effect of temperature and time on the yield of cocoa butter analog. Change in pressure can affect density-dependent physical properties (partition coefficient, dielectric constant, and Hildebrandt solubility parameter) of SC-CO 2, which can indirectly modulate enzyme activity, specificity, and stability [19]. In supercritical conditions, small variations in pressure result in high density values; consequently, it is possible to vary all the other density-dependent properties to a great extent. The increase in density causes an increase in supercritical phase solvent power and, consequently, there is a greater interaction between enzyme and CO 2. Two other factors might contribute to the observed decrease in cocoa butter analog yield with increasing pressure as explained by Nagesha et al. [16]. One factor is the decreased partition rate of substrates in the liquid phase and another is the increased reaction volume. With increased pressure, the solubility of the substrates in the SC-CO 2 phase increases markedly but decreases in the enzyme phase in which the catalysis takes place. This causes dilution of the substrate and results in the retardation of esterification Effect of temperature In the present study, increase in temperature was expected to increases the activity of enzyme in SC-CO 2 and, hence, the yield of cocoa butter analog. However, contour plots (Figs. 3(d) and 4) show that the yield of cocoa butter analog decreased with increase in temperature above 50 C; temperature, therefore, significantly affected enzyme catalysis in SC-CO 2. Temperature also influenced the three dimensional conformation and, hence, the stability of the enzyme. Lipozyme TL IM, however, is a high temperature stable enzyme and can stand temperatures up to 80 C without any change in its activity [20]. At higher temperatures, the decrease in the yield is most likely due to the reduced density of carbon dioxide that caused reduction of solubility of the substrates in SC-CO 2 [21]. Reaction temperature strongly affects the partition of substrates between the SC-CO 2 phase and the enzyme phase as the results of the changes in substrate solubility and the density and viscosity of SC-CO 2. Higher temperatures also contribute to the differential partition of the substrates in SC-CO 2 phase and enzyme phase. Optimal diffusion is beneficial for mass transfer between substrates and products [22,3] Effect of reaction time The contour plots (Figs. 3(a), 4(c), 5(a) and 5(b)) show that, at optimum reaction conditions, there is a steady decrease in the yield of cocoa butter analog with increased incubation time. So, with all the variables at their optimum level, the maximum yield of cocoa butter analog is possible to obtain within incubation time of 3 h. Kim et al. [23] referred to one factor that might contribute to the observed decrease in the yield of cocoa butter analog with increase

6 214 H. Shekarchizadeh et al. / J. of Supercritical Fluids 49 (2009) Fig. 5. The contour plots: (a) mutual effect of time and SSS/CHF ratio, (b) mutual effect of time and water and (c) the mutual effect of water and SSS/CHF ratio on the yield of cocoa butter analog. in reaction time. They reported that longer reaction times resulted in increased acyl migration. With increased acyl migration, the oleic acid appeared at the sn-2 position of TAGs and, consequently, the yield of cocoa butter analog decreased. As oleic acid constitutes about 90% of fatty acids at sn-2 position of TAGs of cocoa butter, so increased acyl migration increases other fatty acids at sn-2 position of TAGs of cocoa butter analog and causes the reduced yield Effect of water Contour maps (Figs. 3(b), 4(a), 5(b) and 5(c)) show the effect of water content with the interaction of other reaction parameters on the yield of cocoa butter analog and its effect is found to be considerable to obtain the maximum yield of cocoa butter analog. To precisely control the water content in the reaction mixture, the lipases used were previously lyophilized under 50 mtorr for 24 h and CHF and SSS were completely dried. Water availability in a reaction system obviously affects the enzyme activity since water may serve not only as a solvent but also as a reactant in hydrolytic reactions. Therefore, water content is considered as an important thermodynamic parameter that represents the enzyme hydration level in a non-aqueous medium, such as SC-CO 2 system, and modulates the equilibrium between hydrolytic and synthetic processes. In the present study, increased concentration of water content gave rise to increased yield of cocoa butter analog in the range of 10 17%. The difference in the water content required could be due to the difference in substrates used and reaction conditions. Presence of water in the reaction system changes/affects the solubility of substrates and products as well as mass transfer resistance [3] Effect of SSS/CHF ratio The contour plots (Figs. 3(c), 4(b), 5(a) and 5(c)) show the effect of SSS/CHF ratio with other reaction parameters at varied levels on the yield of cocoa butter analog. The maximum cocoa butter analog yield was obtained at 1:1 ratio of SSS and CHF with other variables at their optimum values. However, increased water content, shorter incubation time, and lower pressure and temperature at different SSS/CHF ratios gave higher reaction product yields (of cocoa butter analog). With reduced pressure, a higher SSS/CHF ratio is necessary to obtain maximum yield of cocoa butter analog. This is most likely due to the reduced solubility of SSS at lower pressures, so higher contents of SSS should be used [24].

7 H. Shekarchizadeh et al. / J. of Supercritical Fluids 49 (2009) Conclusions Optimization of the production of cocoa butter analog from camel hump fat and tristearin design shows that all the five reaction variables have their effects on the yield of the cocoa butter analog individually and in association with other reaction parameters under SC-CO 2. Especially, the effect of temperature and pressure on the yield of cocoa butter analog is highly pronounced. The predicted model fits well with the experimental results. The pressure, 10 MPa; temperature, 40 C; SSS/CHF ratio, 1:1; water content, 10% (w/w); and incubation time, 3 h were found to be the optimum conditions to achieve the maximum yield of cocoa butter analog. Acknowledgments The authors would like to acknowledge the Isfahan University of Technology for financial support of this research and B. Bahrami for technical assistance. References [1] A. Torbica, O. Jovanovic, B. Pajin, The advantages of solid fat content determination in cocoa butter and cocoa butter equivalents by the Karlshamns method, Eur. Food Res. Technol. 222 (2006) [2] R. Rao, B.R. Lokesh, TG containing stearic acid, synthesized from coconut oil, exhibit lipidemic effects in rats similar to those of cocoa butter, Lipids 38 (2003) [3] K.J. Liu, H.M. Chang, K.M. Liu, Enzymatic synthesis of cocoa butter analog through interesterification of lard and tristearin in supercritical carbon dioxide by lipase, Food Chem. 100 (2007) [4] H. Wang, H. Wu, Ch. Ho, X. Weng, Cocoa butter equivalent from enzymatic interesterification of tea seed oil and fatty acid methyl esters, Food Chem. 97 (2006) [5] T. Hartmann, H.H. Meyer, T. Scheper, The enantioselective hydrolysis of 3- hydroxy-5-phenyl-4-pentenoicacidethylester in supercritical carbon dioxide using lipases, Enzyme Microb. Technol. 28 (2001) [6] I.S.M. Zaidul, N.A.N. Norulaini, A.K.M. Omar, R.L. Smith, Supercritical carbon dioxide (SC-CO 2) extraction and fractionation of palm kernel oil from palm kernel as cocoa butter replacers blend, J. Food Eng. 73 (2006) [7] K. Liu, H. Cheng, R. Chang, J. Shaw, Synthesis of cocoa butter equivalent by lipase-catalyzed interesterification in supercritical carbon dioxide, JAOCS 74 (1997) [8] M.K. Chang, G. Abraham, V.T. John, Production of cocoa butter-like fat from interesterification of vegetable oils, JAOCS 67 (1990) [9] D. Undurraga, A. Markovits, S. Erazo, Cocoa butter equivalent through enzymic interesterification of palm oil midfraction, Process Biochem. 36 (2001) [10] L.W. Khumalo, L. Majoko, J.S. Read, I. Ncube, Characterization of some underutilised vegetable oils and their evaluation as starting materials for lipasecatalysed production of cocoa butter equivalents, Ind. Crop. Prod. 16 (2002) [11] M. Habulin, M. Primozic, Z. Knez, Supercritical fluids as solvents for enzymatic reactions, Acta Chim. Slov. 54 (2007) [12] H. Gunnlaugsdottir, B. Sivik, Lipase-catalyzed alcoholysis of cod liver oil in supercritical carbon dioxide, JAOCS 72 (1995) [13] Z. Knez, M. Habulin, Compressed gases as alternative enzymatic-reaction solvents: a short review, J. Supercrit. Fluids 23 (2002) [14] K. Rezaei, F. Temelli, Lipase-catalyzed hydrolysis of canola oil in supercritical carbon dioxide, JAOCS 77 (2000) [15] I.T. Kadim, O. Mahgoub, R.S. Al-Maqbaly, K. Annamalai, D.S. Al-Ajmi, Effects of age on fatty acid composition of the hump and abdomen depot fats of the Arabian camel (Camelus dromedarius), Meat Sci. 62 (2002) [16] G.K. Nagesha, B. Manohar, K.U. Sankar, Enzymatic esterification of free fatty acids of hydrolyzed soy deodorizer distillate in supercritical carbon dioxide, J. Supercrit. Fluids 32 (2004) [17] Y. Ikushima, N. Saito, M. Arai, H.W. Blanch, Activation of a lipase triggered by interactions with supercritical carbon dioxide in the near-critical region, J. Phys. Chem. 99 (1995) [18] N. Celebi, N. Yildiz, A.S. Demir, A. Calimli, Optimization of benzoin synthesis in supercritical carbon dioxide by response surface methodology (RSM), J. Supercrit. Fluids 47 (2008) [19] S.V. Kamat, B. Iwaskewycz, E.J. Beckman, A.J. Russel, Biocatalytic synthesis of acrylates in supercritical fluids: tuning enzyme activity by changing pressure, Proc. Natl. Acad. Sci. U.S.A. 90 (1993) [20] H. Zhang, X. Xu, J. Nilsson, H. Mu, J. Adler-Nissen, C. Hoy, Production of margarine fats by enzymatic interesterification with silica-granulated Thermomyces lanuginosa lipase in a large-scale study, JAOCS 78 (2001) [21] Sh. Liu, F. Yang, Ch. Zhang, H. Ji, P. Hong, Ch. Deng, Optimization of process parameters for supercritical carbon dioxide extraction of Passiflora seed oil by response surface methodology, J. Supercrit. Fluids 48 (2009) [22] A. Shishikura, K. Fujimoto, T. Suzuki, K. Arai, Improved lipase-catalyzed incorporation of long-chain fatty acids into medium-chain triglycerides assisted by supercritical carbon dioxide extraction, JAOCS 71 (1994) [23] I.H. Kim, S.N. KO, S.M. Lee, S.H. Chung, H. Kim, K.T. Lee, T.Y. Ha, Production of structured lipids by lipase-catalyzed acidolysis in supercritical carbon dioxide: effect on acyl migration, JAOCS 81 (2004) [24] O. Guclu-Ustundag, F. Temelli, Solubility behavior of ternary systems of lipids in supercritical carbon dioxide, J. Supercrit. Fluids 38 (2006)