- Email: [email protected]
Concentration of docosahexaenoic acid (DHA) and eicosapentaenoic acid (EPA) of tuna oil by urea complexation: optimization of process parameters
Journal of Food Engineering 73 (2006) 203–209 www.elsevier.com/locate/jfoodeng
Concentration of docosahexaenoic acid (DHA) and eicosapentaenoic acid (EPA) of tuna oil by urea complexation: optimization of process parameters Shucheng Liu a
a,b,c
, Chaohua Zhang
b,*
, Pengzhi Hong b, Hongwu Ji
b
South China Sea Institute of Oceanology, Chinese Academy of Sciences, Guangzhou, Guangdong Province 510301, China College of Food Science and Technology, Zhanjiang Ocean University, Zhanjiang, Guangdong Province 524025, China c Graduate School of the Chinese Academy of Sciences, Beijing 100049, China
b
Received 13 January 2005 Available online 10 March 2005
Abstract Production of docosahexaenoic acid (DHA) and eicosapentaenoic acid (EPA) concentrates from tuna oil was optimized. In the process, the liquid recovery yield (Y1) and the total content of DHA and EPA (Y2) were response variables, respectively. A three-factor central composite rotatable design (CCRD) was used to study the effect of urea-to-fatty acid ratio (X1), crystallization temperature (X2) and crystallization time (X3). Second order polynomial regression models for Y1 and Y2 were employed to generate response surfaces. The total DHA and EPA (85.02%) and the liquid recovery yield (25.10%) from tuna oil were obtained at a urea-to-fatty acid ratio of 15 (mole/mole), a crystallization temperature of 5 C, and a crystallization time of 20 h. 2005 Elsevier Ltd. All rights reserved. Keywords: Tuna oil; Polyunsaturated fatty acid; Urea complexation; Process optimization
1. Introduction Marine oils are very important for human nutrition and disease prevention because they are rich in polyunsaturated fatty acid (PUFA), especially docosahexaenoic acid (DHA) and eicosapentaenoic acid (EPA). EPA is the precursor of prostaglandins, tromboxanes and leukotrienes, which are effective anti-aggregatory substances (Simopoulos, 1996). DHA is a component of membrane phospholipids of brain and retina cells, consequently is essential for the human health (Simopoulos, 1996). Some studies indicated that the PUFA concentrates, devoid of more saturated fatty acid, are
*
Corresponding author. Tel.:/fax: +86 759 2382049. E-mail address: [email protected] (C. Zhang).
0260-8774/$ - see front matter 2005 Elsevier Ltd. All rights reserved. doi:10.1016/j.jfoodeng.2005.01.020
much better than marine oils themselves since they allow the daily intake of total lipid to be kept as low as possible (Haagsma, Gent, Luten, Jong, & Doorn, 1982). The PUFA concentrates can be produced by several methods, including freezing crystallization, urea complexation, molecule distillation, supercritical fluid extraction, silver ion complexation and lipase concentration. However, the simplest and most efficient technique for obtaining PUFA concentrates in the form of free fatty acids is urea complexation. This is a well-established technique for elimination of saturated and monounsaturated fatty acids (Gamez et al., 2003; Hermann, 1953; Iverson & Weik, 1967; Lucy & Jer, 2001; Strocchi, 1975; Tor & Yi, 2001). Initially the triacylglycerol (TAG) of the oil are split into their constituent fatty acids by alkaline hydrolysis using alcoholic
204
S. Liu et al. / Journal of Food Engineering 73 (2006) 203–209
KOH or NaOH and these free fatty acids are then mixed with an ethanol solution of urea for complex formation. The saturated and monounsaturated fatty acids easily complex with urea and crystallize out on cooling and may subsequently be removed by filtration. The liquid or non-urea complexed fraction is enriched with PUFA. Urea complexation has the advantage that complexed crystals are extremely stable, and filtration does not necessarily have to be carried out at the very low temperatures which solvent crystallization of fatty acids would require. This method is also favored by many researchers because complexation depends upon the configuration of the fatty acid moieties due to the presence of multiple double bonds, rather than pure physical properties such as melting point or solubility (Udaya, Wanasundara, & Shahidi, 1999). The results of one-factor-at-a-time experiments do not reflect actual changes in the environment as they ignore interactions between factors that are present simultaneously. Therefore, in the study, the central composite rotatable design (CCRD) is employed to optimize the reactive conditions to obtain a maximum concentration of DHA and EPA. The method is the preferred experimental design for fitting polynomial models to analyze response surfaces of multi-factor combination. The method of process optimization by response surface methodology (RSM) is a faster and more economical method for gathering research results than classical one-variable-at-a-time or full-factorial experimentation. This method has been successfully adapted in many optimization studies (Namal & Shahidi, 2002; Shieh, Akoh, & Koehler, 1995; Wannasundara & Shahidi, 1996). In the paper, urea complexation of tuna oil was carried out to concentrate DHA and EPA of oil. Variables such as urea-to-fatty acid ratio (mole/mole, X1), crystallization temperature (C, X2) and crystallization time (h, X3) were studied collectively in order to optimize the conditions to obtain a maximum concentration of DHA and EPA.
2.2. Scheme of DHA and EPA concentration from tuna oil by urea complexation The separation of DHA and EPA from the hydrolyzed fatty acids mixture of tuna oil was carried out by urea-fatty acid adducts formation according to the following procedure. Free fatty acids (10 g) were mixed with urea (10%, w/v) in 95% aqueous ethanol and heated at 60 C–70 C with stirring until the whole mixture turned into a clear homogeneous solution. The ratio of urea-to-fatty acids was changing by using different amounts of urea. Initially, the urea-fatty acid adduct was allowed to crystallize at room temperature but other temperature (25 C, 15 C, 0 C, 15 C, 25 C) were used later for different periods for further crystallization. The crystals formed (urea-fatty acid adducts are also referred to as the urea complexing fraction) were separated from the liquid (non-urea complexing fraction) by filtration under suction using a buchner funnel lined with a thin layer of glass wool. The filtrate was diluted with an equal volume of water and acidified to pH 2–3 with 6 mol/L H2SO4; an equal volume of hexane was subsequently added and the mixture was stirred thoroughly, and then transferred to a separator funnel. The hexane layer, containing liberated fatty acids, was separated from the aqueous layer containing urea. The hexane layer was washed with distilled water to remove any remaining urea and then dried over anhydrous sodium sulphate and the solvent was removed using a rotary evaporator. Fatty acids from the crystallization fraction were recovered after addition of water/6 mol/L H2SO4 and hexane in a similar manner. The two fractions were weighed separately and the percentage recovery of each was calculated. The fatty acid compositions of the two fractions were determined using a gas chromatographic procedure.
2.3. Optimization procedure for production of DHA and EPA concentration via urea complexation of tuna oil
2. Materials and methods 2.1. Materials Crude tuna (Thunnus albacares) oil was obtained from tuna head by protease hydrolysis. Refining (R), bleaching (B), and deodorizing (D) of the tuna oil was carried out according to recommended procedures for fish oil (Bimbo, 1998). RBD oil was stored under nitrogen at 20 C until used. The mixture of fatty acids of tuna oil was obtained from tuna oil by the saponification and acidification method, as described elsewhere (Udaya et al., 1999). All other chemicals used in the study were analytical grade.
A three-factor central composition rotatable design (CCRD) was employed to study the responses, such as the liquid recovery yield and the total content of DHA and EPA (Y variables) by urea complexation of tuna oil. The urea-to-fatty acid ratio (X1), crystallization temperature (X2) and crystallization time (X3) were independent variables studied to optimize Y variable (Table 1). Duplicate reactions were carried out at all designed points except at the central point (0, 0, 0) where nine replications were performed to allow the estimation of the ‘‘pure error’’. All experiments were carried out in a randomized order to minimize the effect of unexplained variability in the observed responses due to extraneous factors.
S. Liu et al. / Journal of Food Engineering 73 (2006) 203–209
205
Table 1 Variables (factors) used for CCRD Coded-variables levels (Zj)
X1 (Urea-to-fatty acid ratio, mole ratio)
X2 (Crystallization temperature, C)
X3 (Crystallization time, h)
1.682 1 0 1 1.682
25 20 12.5 5 0
25 15 0 15 25
29 24 16 8 3
X1 = 12.5 + Z1*7.5, X2 = 0 + Z2*15, X3 = 16 + Z3*8.
A quadratic polynomial regression model was assumed for predicting individual Y variables. The model proposed for each response of Y was: X X XX Y ¼ b0 þ bi X i þ bii X 2i þ bij X i X j where b0, bi, bii, bij are intercept, linear, quadratic and interaction regression coefficient terms, respectively, and Xi and Xj are independent variables (Mason, Gunst, & Hess, 1989). The statistical analytical system was used for multiple regression analysis, analysis of variance (ANOVA), canonical analysis and analysis of ridge maximum of data in the response surface regression (RSREG) procedure. Response surfaces and contour plots were developed using the fitted quadratic polynomial equations obtained from RSREG analysis and holding the independent variables with the least effect on the response at a constant value and changing the levels of the other two variables. 2.4. Chemical analysis Peroxide value and iodine value were determined using the IUPAC (Paquot & Hanuntfenne, 1987) method. 2.5. Fatty acid analysis Fatty acid profiles were determined by preparation of methyl esters as described by IUPAC (Paquot & Hanuntfenne, 1987). The fatty acid methyl esters were identified by gas chromatography (Shimadzu GC-14B, Tokyo, Japan) equipped with a flame ionization detector and integrator. A capillary column (30 m · 0.25 mm · 0.25 mm, FFAP) was purchased from Dalian Institute of Chemical Physics, which station phase is polyethylene glycol. The temperature for injector and detector were 260 C. The oven temperature was hold at 190 C for 15 min, then increased to 230 C at 5 C/ min and held at 230 C for 15 min. Nitrogen was used as carrier gas and pressure was 500 kPa. The fatty acid methyl esters were identified by comparison with standards and were quantified as the area percentage of each fatty acid methyl ester. EPA methyl ester and DHA methyl ester standards were purchased from Sigma Chemical Co., St. Louis, USA.
3. Results and discussion Urea-to-fatty acid and crystallization temperature have a very large effect on results of experiments in urea complexation reaction. Generally, enrichment of polyunsaturated fatty acid in concentrate and overall recovery varied inversely with increasing urea-to-fatty acid ratio as well as decreasing crystallization temperature. Therefore, these experimental variables should be carefully controlled in order to achieve a maximum content of DHA and EPA in the concentrate with a reasonable recovery. Some results (Haagsma et al., 1982; Ratnayake, Olsson, Matthews, & Ackman, 1988; Udaya et al., 1999) have reported that DHA is the major portion in the non-urea complexing fraction for urea complexation experiments carried out for cod liver, menhaden oil and seal blubber oil. Although a major portion of EPA was recovered in the non-urea complexing fraction, the content of EPA in the non-urea complexing fraction was less than DHA. On one hand, the content of EPA in the mixture of fatty acid is lower, on the other hand EPA has a greater tendency to form a urea adduct than DHA. Complete removal of saturated fatty acids by urea complexation may be impossible since some of the saturated fatty acids do not complex with urea during crystallization. Long chain monounsaturated fatty acids (MUFA), especially those of the C20 and C22, form complexes with urea more readily than those of shorter chain saturated fatty acids (C14 and C16). 3.1. Analysis of model Analysis of variance of the factors studied for the two response surface model is given Table 2. From statistical analysis, the urea-to-fatty acid ratio was the most important factor because it affected the liquid recovery yield and the total content of DHA and EPA highly significantly. Crystallization temperature affected the liquid recovery yield (Y1) highly significantly, and the total content of DHA and EPA (Y2) significantly. Crystallization time affected the 2 response variables insignificantly. Multiple regression coefficients, obtained by employing a least squares technique to predict a quadratic
206
S. Liu et al. / Journal of Food Engineering 73 (2006) 203–209
Table 2 Analysis variance of the factors studied for the response surface model Response variables
Independent variables
Degree of freedom
Sum of square
Mean square
F-value
P-value
Y1
Z1 Z2 Z3 Z1 Z2 Z3
4 4 4 4 4 4
3233.04 324.31 10.76 5442.15 236.64 23.00
808.26 81.08 2.69 1360.54 59.16 7.00
212.20 21.29 0.71 89.37 3.89 0.46
0.00 0.00 0.61 0.00 0.03 0.76
Y2
Y1: the liquid recovery yield, Y2: the total content of DHA and EPA. P < 0.01 highly significant, 0.01 < P < 0.05 significant, P > 0.05 not significant.
polynomial model for the two response variables, are summarized in Table 3. For the liquid recovery yield, examination of these coefficients with the t-test indicated that linear and quadratic terms of the urea-to-fatty acid ratio and crystallization temperature were highly significant, but crystallization time was insignificant. For the total content of DHA and EPA, linear and quadratic terms of urea-to-fatty acid ratio was highly significant, linear and quadratic terms of crystallization temperature was significant, crystallization time was not significant. There was not significant interaction between any variables tested for the two response variables. Therefore, these results suggest that linear and (or) quadratic effects of the urea-to-fatty acid ratio and crystallization temperature may be the primary determining factors affecting the 2 response variables. The coefficients of independent variables (urea-tofatty acid ratio, crystallization temperature, crystallization time) determined for the quadratic polynomial models for the two response variables are given below:
Table 3 Regression coefficients of predicted quadratic polynomial model for the two response variables Variables
Y 1 ¼ 28:85 11:92Z 1 þ 4:01Z 2 0:09Z 3 þ 8:98Z 21 þ 2:40Z 22 þ 0:68Z 23 1:16Z 1 Z 2 þ 0:28Z 1 Z 3 þ 0:58Z 2 Z 3 Y 2 ¼ 82:78 þ 16:21Z 1 2:88Z 2 0:54Z 3 10:79Z 21 2:64Z 22 1:01Z 23 þ 0:75Z 1 Z 2 þ 0:07Z 1 Z 3 0:98Z 2 Z 3 The analysis of variance and error for the two response surface models are given in Tables 4 and 5. These results show that the models predicted for Y1 and Y2 were adequate as indicated by error analysis that showed nonsignificant lack-of-fit. The regression models for Y1 and Y2 were highly significant with satisfactory coefficients of determination (R2) 0.99 and 0.97, respectively. The models indicated that linear and/or quadratic in the models are the primary determining factors for the two response variables. No statistically significant interaction existed between any two of the three factors. The contribution of linear and quadratic terms to the models was 0.60 and 0.38 for Y1, 0.63 and 0.33 for Y2, respectively. 3.2. Analysis of the stationary point
Coefficients (b) Y1 (%)
Y2 (%)
Intercept
28.85***
82.78***
Linear Z1 Z2 Z3
11.92*** 4.01*** 0.09
16.21*** 2.89** 0.54
Quadratic Z11 Z22 Z33
8.98*** 2.40*** 0.68
10.79*** 2.64** 1.01
Interaction Z12 Z13 Z23 Z123
1.16 0.28 0.58 –
0.75 0.07 0.98 –
R2
0.99
0.97
Y1: the liquid recovery yield, Y2: the total content of DHA and EPA. ***P < 0.01 highly significant, **P < 0.05 significant, no asterisk P > 0.05 not significant.
Canonical analysis was performed on the predicated quadratic polynomial models to examine the overall shape of the response surface curves and used to characterize the nature of the stationary point. Canonical analysis is a mathematical approach used to locate the stationary point of the response surface and to determine whether it represents a maximum, minimum or saddle point. Thus, to determine the nature of the stationary point, canonical analysis was carried out on the second order polynomial model. The canonical forms of the equation demonstrating the nature of the response surface were: Y 1 ¼ 23:74 þ 25:56x21 þ 6:78x22 þ 1:77x23 Y 2 ¼ 89:38 2:49x21 7:81x22 30:57x23 where x1, x2, x3 are the axes of the response surface. It is evident that all the eigenvalues were positive, indicat-
S. Liu et al. / Journal of Food Engineering 73 (2006) 203–209
207
Table 4 Analysis of variance of regression parameters for the response surface model Response variables
Regression
Degree of freedom
Sum of squares
R2
F-value
P-value
Y1
Linear Quadratic Cross product Total
3 3 3 9
2159.13 1374.44 14.10 3547.67
0.60 0.38 0.00 0.99
188.90 120.30 1.23 103.50
0.00 0.00 0.34 0.00
Y2
Linear Quadratic Cross product Total
3 3 3 9
3705.67 1967.66 12.15 5685.49
0.63 0.33 0.00 0.97
81.14 43.08 0.27 41.49
0.00 0.00 0.85 0.00
Y1: the liquid recovery yield, Y2: the total content of DHA and EPA. P < 0.01 highly significant, 0.01 < P < 0.05 significant, P > 0.05 not significant.
Table 5 Analysis of variance for second order polynomial model fitted to response surface Response variables
Source
Degree of freedom
Y1
Lack of fit Pure error Total error
5 8 13
Sum of squares 10.27 39.25 49.52
Mean squares 2.05 4.91 3.81
0.42
F-value
Y2
Lack of fit Pure error Total error
5 8 13
136.26 61.66 197.92
27.25 7.71 27.25
3.54
Y1: the liquid recovery yield, Y2: the total content of DHA and EPA; F0.05 (5, 8) = 3.69, F < 3.69 not significant.
ing that the stationary point was, in fact, a minimum for Y1. All the eigenvalues were negative, indicating that the stationary points were, in fact, a maximum for Y2. The linear, quadratic and cross product terms in the second order polynomial were used to generate a three dimensional response surface graph and a two dimensional contour plot. While the three dimensional response surface graph can assist the researcher to determine the direction to take to increase a desired response and graphically show the nature of the fitted surface as a maximum, minimum or saddle point, it is
difficult to determine the levels of variables from such a graph. This can be more readily achieved from a contour plot of the same variables. In a contour plot, curves of equal response values are drawn on a plane whose coordinates represent the levels of the independent variables. Each contour represents a specific value for the height of the surface, above the plane defined for combination of the levels of the independent variables. Therefore, different surface height values enable one to focus attention on the levels of the independent variables at which changes in the surface height occur.
Fig. 1. Response surface and contour plots for the effect of urea-to-fatty acid ratio and crystallization temperature on the liquid recovery yield of the concentrate of tuna oil.
208
S. Liu et al. / Journal of Food Engineering 73 (2006) 203–209
Fig. 2. Response surface and contour plots for the effect of urea-to-fatty acid ratio and crystallization temperature on the total content of DHA and EPA of the concentrate of tuna oil.
Table 6 Predicted values and Observed values for response variables in urea complexation experiment of tuna oil Response variables
Critical values of independent variables Urea-to fatty acid (mole ratio)
Crystallization temperature (C)
Crystallization time (h)
Stationary point
Predicted value (%)
Observed value (%)
Y1 (%) Y2 (%)
15.23 15.78
6.40 3.88
17.15 15.86
Min Max
23.74 89.38
25.10 85.02
Y1: the liquid recovery yield, Y2: the total content of DHA and EPA.
The stationary point for the total content of DHA and EPA by urea complexation predicted a maximum of 89.38% at a urea-to-fatty acid ratio of 15.78, crystallization temperature of 3.88 C and crystallization time of 15.86 h. The contour plot derived from the results of canonical analysis showed ellipsoidal contours at the maximum point (Fig. 2). But the predicated value of the liquid recovery yield was a minimum (Fig. 1). This shows that the content of DHA and EPA inversely related to the liquid recovery yield. Therefore, these factors must be controlled carefully in order to achieve a suitable result. Lastly, the urea-to-fatty acid ratio of 15, crystallization temperature of 5 C and crystallization time of 20 h may be suitable considering factors involved. The adequacy of the models predicated was examined by performing independent experiments at the optimal conditions. Verification results revealed that the predicated values from these models were reasonably close to observed values (Table 6). Iodine value of the concentrate of DHA and EPA was enhanced from 181.2 to 324.09. But the peroxide value of the concentrate of DHA and EPA was enhanced too; the reason is that polyunsaturated fatty acids were oxidized easily. Therefore, the process must be protected by nitrogen or by adding some appropriate antioxidations.
4. Conclusion The process parameters were found to be: a urea-tofatty acid ratio of 15, a crystallization temperature of 5 C, and a crystallization time of 20 h by the threefactor central composition rotatable design. Verification results revealed that the predicated values from these models were reasonably close to the observed values. Under these conditions, the total content of DHA and EPA can be increased up to 85.02% with a liquid recovery yield of 25.10% of the weight of the original tuna oil by urea complexation. Fish oil production in 2004 is estimated at 640,000 tones, which is about 56% in aquaculture feed, 30% in edible oil, 12% in industry, 2% in pharmaceutical, food and nutrition. Fish oilÕs use in pharmaceutical, food and nutrition is growing at an accelerated pace owing to the wide coverage of fish oilÕs health benefits. The price of highly purity PUFAs is going up, which is higher than the price of refined fish oil. Therefore the economic benefits are very considerable for preparing highly purity PUFAs from fish oil. Urea complexation is a very effective method for concentrating PUFAs from fish oil because it has many advantages that it is simple to operate, the equipment is with little investment, reagent is cheap, urea can be recycled, the production cost is low and it is easy to expand the scale in production for industrialization.
S. Liu et al. / Journal of Food Engineering 73 (2006) 203–209
References Bimbo, A. P. (1998). Guidelines for characterizing food-grade fish oil. Information, 9, 473–483. Gamez, M. N., Noriega, R. J. A., Medina, J. L. A., Ortega, G. J., Monroy, R. J., Toro, V. F. J., Garcia, H. S., & Angulo, O. (2003). Concentration of EPA and DHA from fish oil by hydrolysis and urea complexation. Food Research International, 36, 721–727. Haagsma, N., Gent, C. M., Luten, J. B., Jong, R. W., & Doorn, E. (1982). Preparation of a x-3 fatty acid concentrate from cod liver oil. Journal of America Oil ChemistsÕ Society, 59, 117–118. Hermann, S. (1953). Urea inclusion compounds of fatty acids, progress in the chemistry of fat and other. Lipids, 2, 243–246. Iverson, J. L., & Weik, R. W. (1967). Correlation of fatty acid structure with preferential order of urea complex formation. Journal of the Association of Official Analytical Chemists, 50, 1111–1118. Lucy, S. H., & Jer, H. L. (2001). Fractionation of urea-pretreated squid visceral oil ethyl esters. Journal of America Oil ChemistsÕ Society, 78(5), 473–476. Mason, R. L., Gunst, R. F., & Hess, J. L. (1989). Statistical design and analysis of experiment with applications to engineering and science. New York: John Wiley. Namal, S. S. P. J., & Shahidi, F. (2002). Lipase-catalyzed incorporation of docosahexaenoic acid (DHA) into borage oil: optimization using response surface methodology. Food Chemistry, 77, 115–123.
209
Paquot, C., & Hanuntfenne, A. (1987). IUPAC standard methods for the analysis of oils, fats and derivatives. Blackwell Scientific Publication Ltd., pp. 143–148. Ratnayake, W. M. N., Olsson, B., Matthews, D., & Ackman, R. G. (1988). Preparation of omega-3 PUFA concentrates from fish oils via urea complexation. Fat Science and Technology, 90, 381–386. Shieh, C. J., Akoh, C. C., & Koehler, P. E. (1995). Four factor response surface optimization of the enzymatic modification of triolein to structured lipids. Journal of America Oil ChemistsÕ Society, 72, 619–623. Simopoulos, A. P. (1996). Omega-3 fatty acids. Part I: Metabolic effects of omega-3 fatty acids and essentiality. In G. A. Spiller (Ed.), Handbook of lipids in human nutrition (pp. 56–61). New York: Boca Raton. Strocchi, A. P. (1975). Correlation between urea inclusion compounds and comformational structure of unsaturated C-18 fatty acid methyl esters. Chemical Physical Lipids, 15, 87–94. Tor, C., & Yi, H. (2001). Polyunsaturated fatty acid concentrates from borage and linseed oil fatty acid. Journal of America Oil ChemistsÕ Society, 78(5), 485–488. Udaya, N., Wanasundara, X., & Shahidi, F. (1999). Concentration of omega 3-polyunsaturated fatty acids of seal blubber oil by urea complexation: Optimization of reaction conditions. Food Chemistry, 65, 41–49. Wannasundara, P., & Shahidi, F. (1996). Optimization of hexametaphosphate-assisted extraction flaxseed proteins using response surface methodology. Food Chemistry, 61, 604–607.
Concentration of docosahexaenoic acid (DHA) and eicosapentaenoic acid (EPA) of tuna oil by urea complexation: optimization of process parameters Shucheng Liu a
a,b,c
, Chaohua Zhang
b,*
, Pengzhi Hong b, Hongwu Ji
b
South China Sea Institute of Oceanology, Chinese Academy of Sciences, Guangzhou, Guangdong Province 510301, China College of Food Science and Technology, Zhanjiang Ocean University, Zhanjiang, Guangdong Province 524025, China c Graduate School of the Chinese Academy of Sciences, Beijing 100049, China
b
Received 13 January 2005 Available online 10 March 2005
Abstract Production of docosahexaenoic acid (DHA) and eicosapentaenoic acid (EPA) concentrates from tuna oil was optimized. In the process, the liquid recovery yield (Y1) and the total content of DHA and EPA (Y2) were response variables, respectively. A three-factor central composite rotatable design (CCRD) was used to study the effect of urea-to-fatty acid ratio (X1), crystallization temperature (X2) and crystallization time (X3). Second order polynomial regression models for Y1 and Y2 were employed to generate response surfaces. The total DHA and EPA (85.02%) and the liquid recovery yield (25.10%) from tuna oil were obtained at a urea-to-fatty acid ratio of 15 (mole/mole), a crystallization temperature of 5 C, and a crystallization time of 20 h. 2005 Elsevier Ltd. All rights reserved. Keywords: Tuna oil; Polyunsaturated fatty acid; Urea complexation; Process optimization
1. Introduction Marine oils are very important for human nutrition and disease prevention because they are rich in polyunsaturated fatty acid (PUFA), especially docosahexaenoic acid (DHA) and eicosapentaenoic acid (EPA). EPA is the precursor of prostaglandins, tromboxanes and leukotrienes, which are effective anti-aggregatory substances (Simopoulos, 1996). DHA is a component of membrane phospholipids of brain and retina cells, consequently is essential for the human health (Simopoulos, 1996). Some studies indicated that the PUFA concentrates, devoid of more saturated fatty acid, are
*
Corresponding author. Tel.:/fax: +86 759 2382049. E-mail address: [email protected] (C. Zhang).
0260-8774/$ - see front matter 2005 Elsevier Ltd. All rights reserved. doi:10.1016/j.jfoodeng.2005.01.020
much better than marine oils themselves since they allow the daily intake of total lipid to be kept as low as possible (Haagsma, Gent, Luten, Jong, & Doorn, 1982). The PUFA concentrates can be produced by several methods, including freezing crystallization, urea complexation, molecule distillation, supercritical fluid extraction, silver ion complexation and lipase concentration. However, the simplest and most efficient technique for obtaining PUFA concentrates in the form of free fatty acids is urea complexation. This is a well-established technique for elimination of saturated and monounsaturated fatty acids (Gamez et al., 2003; Hermann, 1953; Iverson & Weik, 1967; Lucy & Jer, 2001; Strocchi, 1975; Tor & Yi, 2001). Initially the triacylglycerol (TAG) of the oil are split into their constituent fatty acids by alkaline hydrolysis using alcoholic
204
S. Liu et al. / Journal of Food Engineering 73 (2006) 203–209
KOH or NaOH and these free fatty acids are then mixed with an ethanol solution of urea for complex formation. The saturated and monounsaturated fatty acids easily complex with urea and crystallize out on cooling and may subsequently be removed by filtration. The liquid or non-urea complexed fraction is enriched with PUFA. Urea complexation has the advantage that complexed crystals are extremely stable, and filtration does not necessarily have to be carried out at the very low temperatures which solvent crystallization of fatty acids would require. This method is also favored by many researchers because complexation depends upon the configuration of the fatty acid moieties due to the presence of multiple double bonds, rather than pure physical properties such as melting point or solubility (Udaya, Wanasundara, & Shahidi, 1999). The results of one-factor-at-a-time experiments do not reflect actual changes in the environment as they ignore interactions between factors that are present simultaneously. Therefore, in the study, the central composite rotatable design (CCRD) is employed to optimize the reactive conditions to obtain a maximum concentration of DHA and EPA. The method is the preferred experimental design for fitting polynomial models to analyze response surfaces of multi-factor combination. The method of process optimization by response surface methodology (RSM) is a faster and more economical method for gathering research results than classical one-variable-at-a-time or full-factorial experimentation. This method has been successfully adapted in many optimization studies (Namal & Shahidi, 2002; Shieh, Akoh, & Koehler, 1995; Wannasundara & Shahidi, 1996). In the paper, urea complexation of tuna oil was carried out to concentrate DHA and EPA of oil. Variables such as urea-to-fatty acid ratio (mole/mole, X1), crystallization temperature (C, X2) and crystallization time (h, X3) were studied collectively in order to optimize the conditions to obtain a maximum concentration of DHA and EPA.
2.2. Scheme of DHA and EPA concentration from tuna oil by urea complexation The separation of DHA and EPA from the hydrolyzed fatty acids mixture of tuna oil was carried out by urea-fatty acid adducts formation according to the following procedure. Free fatty acids (10 g) were mixed with urea (10%, w/v) in 95% aqueous ethanol and heated at 60 C–70 C with stirring until the whole mixture turned into a clear homogeneous solution. The ratio of urea-to-fatty acids was changing by using different amounts of urea. Initially, the urea-fatty acid adduct was allowed to crystallize at room temperature but other temperature (25 C, 15 C, 0 C, 15 C, 25 C) were used later for different periods for further crystallization. The crystals formed (urea-fatty acid adducts are also referred to as the urea complexing fraction) were separated from the liquid (non-urea complexing fraction) by filtration under suction using a buchner funnel lined with a thin layer of glass wool. The filtrate was diluted with an equal volume of water and acidified to pH 2–3 with 6 mol/L H2SO4; an equal volume of hexane was subsequently added and the mixture was stirred thoroughly, and then transferred to a separator funnel. The hexane layer, containing liberated fatty acids, was separated from the aqueous layer containing urea. The hexane layer was washed with distilled water to remove any remaining urea and then dried over anhydrous sodium sulphate and the solvent was removed using a rotary evaporator. Fatty acids from the crystallization fraction were recovered after addition of water/6 mol/L H2SO4 and hexane in a similar manner. The two fractions were weighed separately and the percentage recovery of each was calculated. The fatty acid compositions of the two fractions were determined using a gas chromatographic procedure.
2.3. Optimization procedure for production of DHA and EPA concentration via urea complexation of tuna oil
2. Materials and methods 2.1. Materials Crude tuna (Thunnus albacares) oil was obtained from tuna head by protease hydrolysis. Refining (R), bleaching (B), and deodorizing (D) of the tuna oil was carried out according to recommended procedures for fish oil (Bimbo, 1998). RBD oil was stored under nitrogen at 20 C until used. The mixture of fatty acids of tuna oil was obtained from tuna oil by the saponification and acidification method, as described elsewhere (Udaya et al., 1999). All other chemicals used in the study were analytical grade.
A three-factor central composition rotatable design (CCRD) was employed to study the responses, such as the liquid recovery yield and the total content of DHA and EPA (Y variables) by urea complexation of tuna oil. The urea-to-fatty acid ratio (X1), crystallization temperature (X2) and crystallization time (X3) were independent variables studied to optimize Y variable (Table 1). Duplicate reactions were carried out at all designed points except at the central point (0, 0, 0) where nine replications were performed to allow the estimation of the ‘‘pure error’’. All experiments were carried out in a randomized order to minimize the effect of unexplained variability in the observed responses due to extraneous factors.
S. Liu et al. / Journal of Food Engineering 73 (2006) 203–209
205
Table 1 Variables (factors) used for CCRD Coded-variables levels (Zj)
X1 (Urea-to-fatty acid ratio, mole ratio)
X2 (Crystallization temperature, C)
X3 (Crystallization time, h)
1.682 1 0 1 1.682
25 20 12.5 5 0
25 15 0 15 25
29 24 16 8 3
X1 = 12.5 + Z1*7.5, X2 = 0 + Z2*15, X3 = 16 + Z3*8.
A quadratic polynomial regression model was assumed for predicting individual Y variables. The model proposed for each response of Y was: X X XX Y ¼ b0 þ bi X i þ bii X 2i þ bij X i X j where b0, bi, bii, bij are intercept, linear, quadratic and interaction regression coefficient terms, respectively, and Xi and Xj are independent variables (Mason, Gunst, & Hess, 1989). The statistical analytical system was used for multiple regression analysis, analysis of variance (ANOVA), canonical analysis and analysis of ridge maximum of data in the response surface regression (RSREG) procedure. Response surfaces and contour plots were developed using the fitted quadratic polynomial equations obtained from RSREG analysis and holding the independent variables with the least effect on the response at a constant value and changing the levels of the other two variables. 2.4. Chemical analysis Peroxide value and iodine value were determined using the IUPAC (Paquot & Hanuntfenne, 1987) method. 2.5. Fatty acid analysis Fatty acid profiles were determined by preparation of methyl esters as described by IUPAC (Paquot & Hanuntfenne, 1987). The fatty acid methyl esters were identified by gas chromatography (Shimadzu GC-14B, Tokyo, Japan) equipped with a flame ionization detector and integrator. A capillary column (30 m · 0.25 mm · 0.25 mm, FFAP) was purchased from Dalian Institute of Chemical Physics, which station phase is polyethylene glycol. The temperature for injector and detector were 260 C. The oven temperature was hold at 190 C for 15 min, then increased to 230 C at 5 C/ min and held at 230 C for 15 min. Nitrogen was used as carrier gas and pressure was 500 kPa. The fatty acid methyl esters were identified by comparison with standards and were quantified as the area percentage of each fatty acid methyl ester. EPA methyl ester and DHA methyl ester standards were purchased from Sigma Chemical Co., St. Louis, USA.
3. Results and discussion Urea-to-fatty acid and crystallization temperature have a very large effect on results of experiments in urea complexation reaction. Generally, enrichment of polyunsaturated fatty acid in concentrate and overall recovery varied inversely with increasing urea-to-fatty acid ratio as well as decreasing crystallization temperature. Therefore, these experimental variables should be carefully controlled in order to achieve a maximum content of DHA and EPA in the concentrate with a reasonable recovery. Some results (Haagsma et al., 1982; Ratnayake, Olsson, Matthews, & Ackman, 1988; Udaya et al., 1999) have reported that DHA is the major portion in the non-urea complexing fraction for urea complexation experiments carried out for cod liver, menhaden oil and seal blubber oil. Although a major portion of EPA was recovered in the non-urea complexing fraction, the content of EPA in the non-urea complexing fraction was less than DHA. On one hand, the content of EPA in the mixture of fatty acid is lower, on the other hand EPA has a greater tendency to form a urea adduct than DHA. Complete removal of saturated fatty acids by urea complexation may be impossible since some of the saturated fatty acids do not complex with urea during crystallization. Long chain monounsaturated fatty acids (MUFA), especially those of the C20 and C22, form complexes with urea more readily than those of shorter chain saturated fatty acids (C14 and C16). 3.1. Analysis of model Analysis of variance of the factors studied for the two response surface model is given Table 2. From statistical analysis, the urea-to-fatty acid ratio was the most important factor because it affected the liquid recovery yield and the total content of DHA and EPA highly significantly. Crystallization temperature affected the liquid recovery yield (Y1) highly significantly, and the total content of DHA and EPA (Y2) significantly. Crystallization time affected the 2 response variables insignificantly. Multiple regression coefficients, obtained by employing a least squares technique to predict a quadratic
206
S. Liu et al. / Journal of Food Engineering 73 (2006) 203–209
Table 2 Analysis variance of the factors studied for the response surface model Response variables
Independent variables
Degree of freedom
Sum of square
Mean square
F-value
P-value
Y1
Z1 Z2 Z3 Z1 Z2 Z3
4 4 4 4 4 4
3233.04 324.31 10.76 5442.15 236.64 23.00
808.26 81.08 2.69 1360.54 59.16 7.00
212.20 21.29 0.71 89.37 3.89 0.46
0.00 0.00 0.61 0.00 0.03 0.76
Y2
Y1: the liquid recovery yield, Y2: the total content of DHA and EPA. P < 0.01 highly significant, 0.01 < P < 0.05 significant, P > 0.05 not significant.
polynomial model for the two response variables, are summarized in Table 3. For the liquid recovery yield, examination of these coefficients with the t-test indicated that linear and quadratic terms of the urea-to-fatty acid ratio and crystallization temperature were highly significant, but crystallization time was insignificant. For the total content of DHA and EPA, linear and quadratic terms of urea-to-fatty acid ratio was highly significant, linear and quadratic terms of crystallization temperature was significant, crystallization time was not significant. There was not significant interaction between any variables tested for the two response variables. Therefore, these results suggest that linear and (or) quadratic effects of the urea-to-fatty acid ratio and crystallization temperature may be the primary determining factors affecting the 2 response variables. The coefficients of independent variables (urea-tofatty acid ratio, crystallization temperature, crystallization time) determined for the quadratic polynomial models for the two response variables are given below:
Table 3 Regression coefficients of predicted quadratic polynomial model for the two response variables Variables
Y 1 ¼ 28:85 11:92Z 1 þ 4:01Z 2 0:09Z 3 þ 8:98Z 21 þ 2:40Z 22 þ 0:68Z 23 1:16Z 1 Z 2 þ 0:28Z 1 Z 3 þ 0:58Z 2 Z 3 Y 2 ¼ 82:78 þ 16:21Z 1 2:88Z 2 0:54Z 3 10:79Z 21 2:64Z 22 1:01Z 23 þ 0:75Z 1 Z 2 þ 0:07Z 1 Z 3 0:98Z 2 Z 3 The analysis of variance and error for the two response surface models are given in Tables 4 and 5. These results show that the models predicted for Y1 and Y2 were adequate as indicated by error analysis that showed nonsignificant lack-of-fit. The regression models for Y1 and Y2 were highly significant with satisfactory coefficients of determination (R2) 0.99 and 0.97, respectively. The models indicated that linear and/or quadratic in the models are the primary determining factors for the two response variables. No statistically significant interaction existed between any two of the three factors. The contribution of linear and quadratic terms to the models was 0.60 and 0.38 for Y1, 0.63 and 0.33 for Y2, respectively. 3.2. Analysis of the stationary point
Coefficients (b) Y1 (%)
Y2 (%)
Intercept
28.85***
82.78***
Linear Z1 Z2 Z3
11.92*** 4.01*** 0.09
16.21*** 2.89** 0.54
Quadratic Z11 Z22 Z33
8.98*** 2.40*** 0.68
10.79*** 2.64** 1.01
Interaction Z12 Z13 Z23 Z123
1.16 0.28 0.58 –
0.75 0.07 0.98 –
R2
0.99
0.97
Y1: the liquid recovery yield, Y2: the total content of DHA and EPA. ***P < 0.01 highly significant, **P < 0.05 significant, no asterisk P > 0.05 not significant.
Canonical analysis was performed on the predicated quadratic polynomial models to examine the overall shape of the response surface curves and used to characterize the nature of the stationary point. Canonical analysis is a mathematical approach used to locate the stationary point of the response surface and to determine whether it represents a maximum, minimum or saddle point. Thus, to determine the nature of the stationary point, canonical analysis was carried out on the second order polynomial model. The canonical forms of the equation demonstrating the nature of the response surface were: Y 1 ¼ 23:74 þ 25:56x21 þ 6:78x22 þ 1:77x23 Y 2 ¼ 89:38 2:49x21 7:81x22 30:57x23 where x1, x2, x3 are the axes of the response surface. It is evident that all the eigenvalues were positive, indicat-
S. Liu et al. / Journal of Food Engineering 73 (2006) 203–209
207
Table 4 Analysis of variance of regression parameters for the response surface model Response variables
Regression
Degree of freedom
Sum of squares
R2
F-value
P-value
Y1
Linear Quadratic Cross product Total
3 3 3 9
2159.13 1374.44 14.10 3547.67
0.60 0.38 0.00 0.99
188.90 120.30 1.23 103.50
0.00 0.00 0.34 0.00
Y2
Linear Quadratic Cross product Total
3 3 3 9
3705.67 1967.66 12.15 5685.49
0.63 0.33 0.00 0.97
81.14 43.08 0.27 41.49
0.00 0.00 0.85 0.00
Y1: the liquid recovery yield, Y2: the total content of DHA and EPA. P < 0.01 highly significant, 0.01 < P < 0.05 significant, P > 0.05 not significant.
Table 5 Analysis of variance for second order polynomial model fitted to response surface Response variables
Source
Degree of freedom
Y1
Lack of fit Pure error Total error
5 8 13
Sum of squares 10.27 39.25 49.52
Mean squares 2.05 4.91 3.81
0.42
F-value
Y2
Lack of fit Pure error Total error
5 8 13
136.26 61.66 197.92
27.25 7.71 27.25
3.54
Y1: the liquid recovery yield, Y2: the total content of DHA and EPA; F0.05 (5, 8) = 3.69, F < 3.69 not significant.
ing that the stationary point was, in fact, a minimum for Y1. All the eigenvalues were negative, indicating that the stationary points were, in fact, a maximum for Y2. The linear, quadratic and cross product terms in the second order polynomial were used to generate a three dimensional response surface graph and a two dimensional contour plot. While the three dimensional response surface graph can assist the researcher to determine the direction to take to increase a desired response and graphically show the nature of the fitted surface as a maximum, minimum or saddle point, it is
difficult to determine the levels of variables from such a graph. This can be more readily achieved from a contour plot of the same variables. In a contour plot, curves of equal response values are drawn on a plane whose coordinates represent the levels of the independent variables. Each contour represents a specific value for the height of the surface, above the plane defined for combination of the levels of the independent variables. Therefore, different surface height values enable one to focus attention on the levels of the independent variables at which changes in the surface height occur.
Fig. 1. Response surface and contour plots for the effect of urea-to-fatty acid ratio and crystallization temperature on the liquid recovery yield of the concentrate of tuna oil.
208
S. Liu et al. / Journal of Food Engineering 73 (2006) 203–209
Fig. 2. Response surface and contour plots for the effect of urea-to-fatty acid ratio and crystallization temperature on the total content of DHA and EPA of the concentrate of tuna oil.
Table 6 Predicted values and Observed values for response variables in urea complexation experiment of tuna oil Response variables
Critical values of independent variables Urea-to fatty acid (mole ratio)
Crystallization temperature (C)
Crystallization time (h)
Stationary point
Predicted value (%)
Observed value (%)
Y1 (%) Y2 (%)
15.23 15.78
6.40 3.88
17.15 15.86
Min Max
23.74 89.38
25.10 85.02
Y1: the liquid recovery yield, Y2: the total content of DHA and EPA.
The stationary point for the total content of DHA and EPA by urea complexation predicted a maximum of 89.38% at a urea-to-fatty acid ratio of 15.78, crystallization temperature of 3.88 C and crystallization time of 15.86 h. The contour plot derived from the results of canonical analysis showed ellipsoidal contours at the maximum point (Fig. 2). But the predicated value of the liquid recovery yield was a minimum (Fig. 1). This shows that the content of DHA and EPA inversely related to the liquid recovery yield. Therefore, these factors must be controlled carefully in order to achieve a suitable result. Lastly, the urea-to-fatty acid ratio of 15, crystallization temperature of 5 C and crystallization time of 20 h may be suitable considering factors involved. The adequacy of the models predicated was examined by performing independent experiments at the optimal conditions. Verification results revealed that the predicated values from these models were reasonably close to observed values (Table 6). Iodine value of the concentrate of DHA and EPA was enhanced from 181.2 to 324.09. But the peroxide value of the concentrate of DHA and EPA was enhanced too; the reason is that polyunsaturated fatty acids were oxidized easily. Therefore, the process must be protected by nitrogen or by adding some appropriate antioxidations.
4. Conclusion The process parameters were found to be: a urea-tofatty acid ratio of 15, a crystallization temperature of 5 C, and a crystallization time of 20 h by the threefactor central composition rotatable design. Verification results revealed that the predicated values from these models were reasonably close to the observed values. Under these conditions, the total content of DHA and EPA can be increased up to 85.02% with a liquid recovery yield of 25.10% of the weight of the original tuna oil by urea complexation. Fish oil production in 2004 is estimated at 640,000 tones, which is about 56% in aquaculture feed, 30% in edible oil, 12% in industry, 2% in pharmaceutical, food and nutrition. Fish oilÕs use in pharmaceutical, food and nutrition is growing at an accelerated pace owing to the wide coverage of fish oilÕs health benefits. The price of highly purity PUFAs is going up, which is higher than the price of refined fish oil. Therefore the economic benefits are very considerable for preparing highly purity PUFAs from fish oil. Urea complexation is a very effective method for concentrating PUFAs from fish oil because it has many advantages that it is simple to operate, the equipment is with little investment, reagent is cheap, urea can be recycled, the production cost is low and it is easy to expand the scale in production for industrialization.
S. Liu et al. / Journal of Food Engineering 73 (2006) 203–209
References Bimbo, A. P. (1998). Guidelines for characterizing food-grade fish oil. Information, 9, 473–483. Gamez, M. N., Noriega, R. J. A., Medina, J. L. A., Ortega, G. J., Monroy, R. J., Toro, V. F. J., Garcia, H. S., & Angulo, O. (2003). Concentration of EPA and DHA from fish oil by hydrolysis and urea complexation. Food Research International, 36, 721–727. Haagsma, N., Gent, C. M., Luten, J. B., Jong, R. W., & Doorn, E. (1982). Preparation of a x-3 fatty acid concentrate from cod liver oil. Journal of America Oil ChemistsÕ Society, 59, 117–118. Hermann, S. (1953). Urea inclusion compounds of fatty acids, progress in the chemistry of fat and other. Lipids, 2, 243–246. Iverson, J. L., & Weik, R. W. (1967). Correlation of fatty acid structure with preferential order of urea complex formation. Journal of the Association of Official Analytical Chemists, 50, 1111–1118. Lucy, S. H., & Jer, H. L. (2001). Fractionation of urea-pretreated squid visceral oil ethyl esters. Journal of America Oil ChemistsÕ Society, 78(5), 473–476. Mason, R. L., Gunst, R. F., & Hess, J. L. (1989). Statistical design and analysis of experiment with applications to engineering and science. New York: John Wiley. Namal, S. S. P. J., & Shahidi, F. (2002). Lipase-catalyzed incorporation of docosahexaenoic acid (DHA) into borage oil: optimization using response surface methodology. Food Chemistry, 77, 115–123.
209
Paquot, C., & Hanuntfenne, A. (1987). IUPAC standard methods for the analysis of oils, fats and derivatives. Blackwell Scientific Publication Ltd., pp. 143–148. Ratnayake, W. M. N., Olsson, B., Matthews, D., & Ackman, R. G. (1988). Preparation of omega-3 PUFA concentrates from fish oils via urea complexation. Fat Science and Technology, 90, 381–386. Shieh, C. J., Akoh, C. C., & Koehler, P. E. (1995). Four factor response surface optimization of the enzymatic modification of triolein to structured lipids. Journal of America Oil ChemistsÕ Society, 72, 619–623. Simopoulos, A. P. (1996). Omega-3 fatty acids. Part I: Metabolic effects of omega-3 fatty acids and essentiality. In G. A. Spiller (Ed.), Handbook of lipids in human nutrition (pp. 56–61). New York: Boca Raton. Strocchi, A. P. (1975). Correlation between urea inclusion compounds and comformational structure of unsaturated C-18 fatty acid methyl esters. Chemical Physical Lipids, 15, 87–94. Tor, C., & Yi, H. (2001). Polyunsaturated fatty acid concentrates from borage and linseed oil fatty acid. Journal of America Oil ChemistsÕ Society, 78(5), 485–488. Udaya, N., Wanasundara, X., & Shahidi, F. (1999). Concentration of omega 3-polyunsaturated fatty acids of seal blubber oil by urea complexation: Optimization of reaction conditions. Food Chemistry, 65, 41–49. Wannasundara, P., & Shahidi, F. (1996). Optimization of hexametaphosphate-assisted extraction flaxseed proteins using response surface methodology. Food Chemistry, 61, 604–607.