Supercritical Carbon Dioxide Extraction Efficiency for Carotenes from Carrots by RSM

Supercritical Carbon Dioxide Extraction Efficiency for Carotenes from Carrots by RSM P.J. VEGA, M.O. BALABAN, C.A. SIMS, S.F. O’KEEFE and J.A. CORNELL

ABSTRACT Increasing demand for b-carotene has resulted in a growing interest in its extraction from natural sources. Carotenes were extracted from carrot pulp (press cake) using SC-CO2. Three levels of pressure (20.7, 27.6, and 34.5 MPa), temperature (40, 55, and 707C), and ethanol co-solvent (0, 5, 10% wt) were studied. Percentages a-, b- and total carotenes extracted were determined by HPLC and spectrophotometric methods. A maximum of 99.5% of b-carotene was extracted using 10% ethanol. Concentration of ethanol and temperature were the most important factors in increasing extraction yield. A response surface model was developed from the results: %b-carotene extraction 5 8.4558 1 0.8816 T 2 10.7650 E 1 0.1843 T*E 1 1.9019 E2 2 0.0261 E2*T, where T 5 temperature (7C), E 5 ethanol (%). R2 5 0.973. Key Words: carotene, carrot, supercritical CO2 extraction, Response Surface Methodology

INTRODUCTION INCREASING DEMAND for natural b-carotene (Anonymous, 1992) has resulted in growing interest in extracting b-carotene from vegetable products. Besides their pro-vitamin A activity, a- and b-carotene as antioxidants (Graham, 1988) in the diet may have important effects in cancer prevention (Shekelle et al., 1981; Krinsky, 1988) and reduction of the risk of heart disease (Kardinaal et al., 1993). Carotenes are used commercially as natural food colorants. They are responsible for many red, orange and yellow colors of edible fruits, vegetables and mushrooms. They have been added to many products like butter, popcorn, salad dressings and beverages (Gordon and Bauerfeind, 1983). Carotenes in the market are synthetic b-carotene and natural carotene extracted from the mass cultivated algae Dunaliella salina and D. bardawilli. Extraction of carotenes from these algae involves use of organic solvents such as hexane (Anonymous, 1992). Total carrot production in the U.S. during 1991 was estimated at 1.4 million tons (Agricultural Statistics Board, 1992). Carrots contain up to 500 ppm of total carotenoids, and b-carotene accounts for 45% to 65%, a-carotene 20% to 40%, Z-carotene 5% to 10% and others up to 5% of that amount (Phillip et al., 1989). A notable portion of total carrot production does not reach the fresh market due to size or shape defects. Some of these are used for production of carrot juice. Pressing is effective in obtaining high juice yields, but not in extraction of carotenes. Our preliminary research indicated that ,30% of the b-carotene was extracted by pressing ground carrots in a hydraulic press. Thus, a high percentage of a valuable component remained in the press cake, which has not been utilized. Extraction of carotenoids from carrot press cake could enhance the value of carrots and could create a natural, high value pigment and potential nutritional supplement from this underutilized by-product. Most extraction methods of carotenes utilize organic solvents such as hexane (Sadler et al., 1990; Mosher, 1961). This involves the direct cost of the solvent as well as the need to remove all potentially toxic solvent residues.

The authors are affiliated with the Food Science & Human Nutrition Dept., Univ. of Florida, Gainesville, FL 32611.

Several researchers have noted that supercritical CO2 (SCCO2) extraction of carotenoids may be an effective alternative to the traditional solvent extraction (Chao et al., 1991; Lorenzo et al., 1991; Favati et al., 1988). Lorenzo et al. (1991) reported SC-CO2 extraction of b-carotene from algae, but did not indicate efficiency. Favati et al. (1988) reported that .90% of the carotene content of freeze dried leaf protein concentrate was extracted by SC-CO2. Using SC-CO2 Yamaguchi et al. (1986) extracted astaxanthin from antarctic krill, and Degnan et al. (1991) extracted bixin (a carotenoid-type pigment) from annatto seeds. Chao et al., (1991) indicated that a higher extraction of pigments from annatto seed was obtained when using pressures .31.0 MPa than at 21.0 MPa. Spanos et al. (1993) extracted 80% of the total b-carotene present in freeze-dried sweet potato. Sample moisture content can be critical in carotene extraction, presence of moisture (about 50%, compared to freeze-dried samples) increased extraction of astaxanthin from crawfish shells (Charest, 1993). Polar compounds such as ethanol have improved extraction efficiency (Temelli, 1992; Charest, 1993). Cygnarwicz and Seider (1990b) reported that with high operating pressures, SC extraction could still be competitive for recovery of high-value products at low production rates. A complete carotene extraction from fresh samples of carrot was obtained by Goto et al. (1993) using SC-CO2. Independent effects of pressure, temperature, ethanol and moisture content were studied. A systematic quantification of the effects of these factors and their interactions is needed through a surface response model to develop optimum conditions. Our objectives were, (1) to investigate the effects of temperature, pressure, and % ethanol on the extraction of carotenes from carrots; and, (2) to develop a prediction equation for total carotenes extracted with different temperatures, pressures, and ethanol concentrations. MATERIALS & METHODS Experimental design A common experimental design for investigating linear and quadratic effects of two or more factors is the Box-Behnken (1960) design, where each factor is varied over three levels. For the three factors (temperature, pressure, and ethanol), a 13-point Box-Behnken design was used. The standard error for the whole experiment was estimated from the standard error obtained from replications of the center point conditions. The amounts of carotene extracted at the 13 factor-level conditions were analyzed by fitting regression equations. Estimated percent extracted a-, b-, and total carotenes surfaces were plotted (Cornell, 1984). Sample preparation Carrot (cv. Apache) press cake was obtained from Florida Food Products Inc., Eustis, Florida. The procedure involved first rinsing raw carrots, then grinding the carrots in a hammer mill, heating the ground carrots to 907C for '10 min, cooling to '607C and pressing with a ‘‘Squeeze Box’’ type press. After juice was extracted, the press cake was used for the study. 11.4 kg of a 81 52% moisture content pulp from the same press batch were collected, hand mixed for 15 min and placed in 50g sealable plastic pouches. The product was immediately frozen using solid CO2, and transported frozen to the Food Science & Human Nutrition Dept., Univ. of Florida, Gainesville, Fl., where the frozen samples were kept in the dark at 2207C until needed. The average

Volume 61, No. 4, 1996—JOURNAL OF FOOD SCIENCE—757

SC EXTRACTION OF b-CAROTENE . . . PA) was used to monitor total volume (250L STP), and flow rate (1.5 L/min approx.) of CO2, which was kept constant for all treatments. CO2 flow through the vessel was stopped when the totalizer volume reached 250L. This did not include CO2 that came out of the vessel during depressurization. In extractions with 5 and 10% ethanol, the extract contained about 25 mL and 50 mL of ethanol, respectively. The ethanol was evaporated using a Rotary Evaporator, model 50–60 CY (Buchler Instruments. Fort Lee, N.J) at 607C and 740 mm Hg of vacuum. Carotene determination

Fig. 1—Flow diagram of supercritical extraction system.

Table 1—Percentage carotenes extracted under the different conditions % carotene extracted

Temperature (&C)

Pressure (MPa)

Ethanol (%wt)

a-a

b-a

Totalb

70 55 55 40 70 70 55 55 55 40 40 70 55 55 40

27.6 34.5 20.7 27.6 34.5 20.7 27.6 27.6 27.6 34.5 20.7 27.6 34.5 20.7 27.6

10 10 10 10 5 5 5 5 5 5 5 0 0 0 0

96.70 99.10 99.27 93.77 74.76 87.00 62.34 62.55 71.53 47.12 47.73 68.31 58.16 54.46 41.69

96.70 99.51 99.10 93.34 78.56 87.48 63.05 62.53 69.35 46.14 50.90 68.17 62.83 55.07 41.72

97.00 98.81 98.92 94.60 79.84 88.55 63.69 59.64 77.14 52.75 57.70 76.68 55.84 58.03 44.73

a Determined by HPLC. b Determined by Spectrophotometer.

diameter of the particles was 0.93 mm., and the largest particles measured '3.5 mm in diam. Prior to each extraction, one plastic pouch was taken from the freezer and allowed to thaw, from which two 6g samples were weighed. One sample was used for SC-CO2 extraction and the other sample was used as a control for carotene determination. Supercritical carbon dioxide extraction Previous research by Spanos et al. (1993) suggested that higher yields were obtained with higher pressures. Considering the maximum working pressure for our equipment, the three levels of pressure we studied were 20.7, 27.6, and 34.5 MPa. Temperature increases the solubility of bcarotene (Sakaki, 1992), but can also affect degradation (Marty and Berset, 1990). We used the three levels 40, 55, and 707C. As a co-solvent, ethanol was studied at three levels, 0, 5, and 10% by weight. A Milton Roy SC extractor (Model X-10, Milton Roy, Ivyland, PA.) was used. Its maximum operating pressure was around 34.5 MPa, and it had a 150 mL nominal stainless steel pressure vessel heated with an electrical jacket. Thermocouples were placed inside the extraction vessel and between the jacket and the vessel to control temperature. The system had a backpressure regulator valve that could be used to control pressure. For each experimental trial, a 6g sample was placed in a stainless steel mesh cylinder (0.4 mm opening). This container fitted snugly into the extraction vessel. Carrot cake was not compressed in the sample holder, since previous experiments showed that extraction of pigments occurred only from the outer layers of compressed pulp. The sample holder was placed in the pre-heated extraction vessel, and CO2 (Liquid Carbonic, Chicago, IL.) or a mixture of CO2 and ethanol provided by Liquid Carbonic (Chicago, IL.) was pumped through the sample. After the SC fluid passed through the vessel, it expanded and continued flowing through a cold trap (14/20 Alihn condenser, Kontes Specialty Glassware, Vineland, NJ) connected to a 100 mL flask with a side tube to collect the extract (Fig. 1). A flow totalizer (Singer American Meter Division, Philadelphia,

The total, a- and b-carotene content in the original sample, SC extract and residue were determined using the HPLC procedure of Bushway and Wilson (1982). A spectrophotometric method was used to confirm results. The following modifications were used: due to the presence of ethanol and loss of water in some treatments, all samples were dried using a Model 5831 (National Appliance Co., Portland, OR) vacuum oven set at 740 mm Hg vacuum and 727C for 3.5 hr. Preliminary studies compared carotene contents from air dried, freeze-dried and fresh carrot pulp and showed no significant loss by air drying. For HPLC analysis, 0.4g of dry sample were used in duplicate experiments. The dry samples were ground using a mortar and pestle. Due to the small quantities used for analysis, a Brinkmann Tissue Homogenizer (Brinkmann Instruments, Inc. Westburry, NY) was used. The extracts were diluted with HPLC grade tetrahydrofuran to different volumes such that the concentration fell inside the standard curve range. The standard curve was determined after injecting 0.025 to 0.25 µg of pure a-carotene (Sigma Chemical Co.) and 0.05 to 0.5 µg of pure b-carotene (Sigma Chemical Co.). The HPLC system consisted of a Waters 501 HPLC pump, a Nova-Pak C18, ˚ , 4 µm, 3.9*150 mm column (Millipore Corporation, Milford, MA), 60A a U 6K injector, a UV-VIS detector, and a Millenium 2010 data processor integrator. The total carotenoid content was determined using the same extraction procedure by diluting the extract to an appropriate concentration and reading its absorbance at 470 nm in a Perkin-Elmer UV/VIS Spectrophotometer. Results were based on a pure b-carotene (Sigma Chemical Co.) standard curve. Analysis of results Total or 100% recovery of the SC extracted carotenes was not possible since some extract was lost on the pipes, valves, vessel surfaces etc. The true amount extracted was calculated by taking the difference of carotenes initially present in the sample, minus those that remained in the pulp after SC extraction. We assumed that there was minimal or no degradation of carotenes during SC-CO2 extraction. This assumption was valid since light, oxygen and high temperatures were not present in the process (Chao et al., 1991; Lorenzo et al., 1991). To study effects of temperature (T), pressure (P), and ethanol (E) on the % carotene extracted, the data were analyzed by fitting multiple regression equations using Statistica (StatSoft. Tulsa, OK). The initial regression equation fitted was a 10-term third-degree polynomial in T, P, and E. When pressure was discovered to have no effect on % carotene extracted, a reduced form of third-degree polynomial in T and E proved adequate. Percent carotene extracted surface plots were generated from the final regression equations to illustrate the effects of temperature and ethanol (Cornell, 1984).

RESULTS & DISCUSSION THE INITIAL TOTAL b-CAROTENE CONTENT in the carrot was 673 5 200 µg/g fresh pulp. The variation was due to deviations in the HPLC readings. On a given day, there were differences in levels read by HPLC compared to other days. However, replications of the same sample during the same day showed no significant deviation. Therefore, we calculated percent extraction yield rather than absolute amounts since this eliminated effects of errors in absolute readings. Results for % yields for a, band total carotenes were similar (Table 1). This was expected due to the similarity of a- and b-carotene and the other carotenoids in the carrot pulp. Temperature (T) and ethanol (E) were the only significant factors affecting % carotene extraction, with pressure having little or no effect. The effects of T and E on % carotene extraction were modeled using the reduced form of third-degree polynomial.

758—JOURNAL OF FOOD SCIENCE—Volume 61, No. 4, 1996

Table 2—Coefficient estimates and R2 values for fitted regression equations, % carotene extracted 5 b0 1 b1T 1 b2E 1 b3T*E 1 b4E2 1 b5T*E2 Coefficient estimates

% Carotene extracted

b0

b1

b2

b3

b4

b5

R2

abTotal

6.852 8.456 25.142

0.887 0.882 1.162

29.886 210.765 0.010

0.170 0.184

1.838 1.902 0.962

20.025 20.026 20.011

0.9719 0.9732 0.9458

Fig. 2—Predicted response surface and experimental percentage SC-CO2 extracted b-carotene. Experimental values include the three different pressures used.

Response 5 b0 1 b1T 1 b2E 1 b3T*E 1 b4E2 1 b5T*E2 1 ∈

(1)

In the model the terms b1T and b2E represent linear effects of temperature and ethanol, b4E2 represents the quadratic effect of ethanol, and b3T*E and b5T*E2 represent the linear by linear, and linear by quadratic interaction effects of temperature and ethanol, respectively, on the response. The model fitting consisted of testing the significance of the coefficient estimates of the parameters b1, b2, . . . . , b5 in (1) and dropping the terms b3T*E, b5E2, and b5T*E2 that were significant at p ≤ 0.05. When terms were dropped, the simpler, reduced model was refitted and the coefficient estimate retested. The final model forms contained the first three terms in (1) plus any or all of the last three terms in (1) that were significant at p ≤ 0.05. The coefficient estimates associated with the final fitted regression equations for a-, b-, and total percent carotene extracted were compared (Table 2) along with the values of coefficients of determination (R2) for each model. Temperature increase had a positive effect on the response (% b-carotene extracted) (Fig. 2) but the effect diminished with increasing level of ethanol. This was the reason for the significant interaction term b3T*E. The higher the % ethanol present, the less was the effect of increasing temperature. Ethanol had a positive quadratic effect on the response but mainly at low and mid levels of temperature. At higher temperatures, the positive effect of ethanol was linear. This was the reason for the significant interaction term b5T*E2. This term implies the quadratic effect of ethanol differed linearly across the range of temperature. Low temperature would have a definite quadratic ethanol effect, high temperature a linear ethanol effect. From the shape of the surface (Fig. 2), it was apparent that high (.90%) % b-carotene extraction is possible with 10% ethanol, regardless of temperature. Increasing pressure increases SC-CO2 density, which should improve extraction efficiency. A higher temperature would de-

crease SC-CO2 density, but the solute vapor pressure would increase, thus increasing the solubility of carotene. The increase in solubility had a stronger effect than the increase of density caused by the pressure increase. Other studies of extraction of b-carotene from vegetables, like that of Spanos (1993), showed an interaction effect between temperature and pressure, where effects of pressure were maximized by increased temperature. Our results do not show significant interactive effects, which may have been due to the range of pressure used, which corresponded to the cross-over area of the interactions. In order to determine if there was an interaction between pressure and temperature, pressures lower and higher than used in this study should be included. Ethanol concentration in the CO2 was an important factor contributing to the increase in extraction, as had been reported (Brunner and Peter, 1982; Charest 1993; Cygnarowics et al., 1990a). Solubility of materials with low volatility is increased when using an entrainer compared to SC gas alone (Brunner and Peter, 1982). The high temperature dependency and low pressure dependency in our study could also be explained by the presence of ethanol. The two main effects of ethanol are solubility enhancement and temperature dependance of such solubility enhancement (Brunner and Peter, 1982). Studies by Sakaki (1992) showed that the increase of temperature had a stronger effect on solubility than did the increase of pressure. In a different solubility study (Cygnarowicz et al., 1990a) at 40, 60, and 707C and at pressures between 200 and 500 bar, most favorable conditions were at 707C and 439 bar (43.9 MPa). At the optimum conditions, addition of 1% ethanol increased solubility over 100%. The amount of extracted recovered carotene was low mainly due to the relatively small amount of carotene present in 6g of carrot pulp. Recovery was defined as the percent of extract obtained from the experiment, compared to the theoretical yield calculated as the difference of carotenes in the initial sample and the SC extracted pulp. As expected, higher recoveries were obtained by increasing the percentage ethanol. Ethanol facilitated recovery by carrying the carotenes after being extracted from the carrot pulp. Extractions with no ethanol resulted in ' 30 2 40% recovery of extracts, while 5 and 10% ethanol resulted in 50–65% recoveries. CONCLUSION OBTAINING HIGH EFFICIENCY EXTRACTIONS of b-carotene from an under-utilized natural source (carrot pulp) is feasible by SC-CO21 ethanol extraction. The press cake after expression of juice from carrots could be used directly for supercritical CO2 extraction, with no particle size reduction or drying needed. The high value of natural b-carotene, the availability of low cost carrot pulp byproduct, and the high extraction yield that could be achieved suggest the positive economic feasibility of the process. REFERENCES Agricultural Statistics Board. 1992. Vegetables—1991 Preliminary Acreage, Yield and Production. National Agricultural Statistics Service, USDA: 26–27. Anonymous. 1992. Palm oil yields carotene for world markets. Inform 3(2): 210–217. Box, G.E.P. and Behnken, D.W. 1960. Some new three level designs for the study of quantitative variables. Technometrics 2(4): 455–476. Brunner, G. and Peter, S. 1982. On the solubility of glycerides and fatty acids in compressed gases in the presence of an entrainer. Separation Sci. & Technol. 17(1): 199–214.

—Continued on page 765

Volume 61, No. 4, 1996—JOURNAL OF FOOD SCIENCE—759