Separation of phosphatidylcholine from soybean phospholipids by simulated moving bed

Separation of phosphatidylcholine from soybean phospholipids by simulated moving bed first-author †E-mail:lvyubin2001@zju.edu.cnwupd@zju.edu.cn †E-mail:lvyubin2001@zju.edu.cnwupd@zju.edu.cn lvyubin2001@zju.edu.cn corresp ‡Corresponding Author ‡Corresponding Author †E-mail:lvyubin2001@zju.edu.cnwupd@zju.edu.cn †E-mail:lvyubin2001@zju.edu.cnwupd@zju.edu.cn lvyubin2001@zju.edu.cn National Laboratory of Secondary Resources Chemical Engineering, Zhejiang University, Hangzhou 310027, China National Laboratory of Secondary Resources Chemical Engineering, Zhejiang University, Hangzhou 310027, China A simulated moving bed (SMB), equipped with eight silica-gel columns, was used to separate phosphatidylcholine (PC) from soybean phospholipids. The effects of flow rate in Sections 2 ( Q 2 ) and 3 ( Q 3 ), switching time, feed flow rate and feed concentration on the operating performance parameters: purity, recovery, productivity and desorbent consumption were studied. Operating conditions leading to more than 90% purity in both outlet streams have been identified, together with those achieving optimal performance. Regions leading to complete separation are observed and explained theoretically. As the mass-transfer effect was not considered, the triangle theory only gives initial guesses for the optimal operating conditions. Simulated moving bed (SMB) Separation Soybean phosphatidylcholine (PC) INTRODUCTION 1 Natural phospholipids, main constituents of cell membranes, have many important biological functions for all living organisms. High-purity phosphatidylcholine (PC), a fraction of phospholipids, has been extensively applied as emulsification, stabilization and wetting agents in the fields of health products, pharmaceuticals, foods and industrial manufacture of various products, etc. High-purity PC prepared by preparative chromatography had been reported in (Li, 2004 2004 ; Amari and Brown, 1990 1990 ). Its inherent shortcomings are low loading amount and high desorbent consumption, which lead to high investment and production cost. Simulated moving bed (SMB) is a continuous chromatographic countercurrent process and has been widely used on large scale in the past few decades (Ruthven, 1984 1984 ; Ruthven and Ching, 1989 1989 ). Compared to preparative chromatography, the adsorbent inventory and the desorbent consumption are much less. In addition, high performance can be achieved even when the adsorbent selectivity components is rather low (Jupke et al., 2002 2002 ; Bruce, 1998 1998 ). It is known that the affinity of PC to the stationary phase (silica gel) is higher than that of all the other phospholipids in the soybean crude lipids (Meeren et al., 1990 1990 ). Therefore, SMB is expected to be a feasible process for separation/purification of PC from crude lipids. This work was aimed at evaluating the SMB performance in preparation of high purity PC and at studying the effect of operation parameters (flow rates in each section of SMB, switching time, feed flow rate and feed concentration) on purity, recovery of PC, productivity, and desorbent consumption. It was expected that optimal operation conditions could be obtained from this study and could be useful for scaling up of a separation process of PC by SMB. EXPERIMENTAL DETAILS 1 Apparatus 2 The SMB unit used was a CSEP 9116 from Knauer, Berlin (Fig. 1 1 ). It consisted of a carousel of 8 stainless steel columns (100 mm×20 mm i.d.) connected to a single multi-function valve. Two HPLC pumps (K-501, Knauer) pumped the feed and the eluent and another two pumps partially recycled the inner currents from Section 1 into Section 2 and from Section 3 into Section 4. The temperature of the columns was maintained at 30 °C by enclosing the entire rotating system within a thermostatically controlled system. The stationary phase was silica gel (20~30 μm), purchased from Meigao Chemical Group (Qingdao, China), where Brunauer-Emmett-Teller (BET) surface is: 312 m 2 /g, average pore diameter is 11.8 nm. The mobile phase was n -hexane-isopropanol-water (1:2.1:0.5, v/v/v). The composition of the samples from SMB was determined by HPLC with column of 100 mm×4.6 mm i.d., packed with silica gel (5 μm). The mobile phase was ethanol-water (80:20, v/v), at flow rate of 1.0 ml/min. The wavelength of the UV detector was set as 205 nm. Phosphatidylcholine purchased from Lipoid (purity>98%, Germany) was used as standard for quantitative determination. Column preparation and characterization 2 A slurry-packing technique was used for packing the columns. The total porosity of the column was determined by the retention volume of a pulse of 1,3,5-tri-tert-butyl-benzene. The bed voidage was calculated from Table 1 1 showing the parameters of the eight columns and their performance. No. Dead time (min) Total porosity Deviation (%) 1 3.640 0.580 0.49 2 3.603 0.574 0.52 3 3.628 0.578 0.17 4 3.654 0.582 0.88 5 3.604 0.574 0.50 6 3.679 0.586 1.57 7 3.539 0.564 2.30 8 3.629 0.578 0.19 Average 3.622 0.577 Adsorption isotherms 2 Since the concentration of soybean lipids in the SMB feed was quite low, linear adsorption isotherm was applied. A pulse of diluted mixture was injected into a column and the retention times of PC peaks and other components V R, i were determined. The Henry constant K i was calculated using the following equation: $$K_i {\text{ = }}\left( {\frac{{V_{{\text{R}},i} - V_0 }}{{V_0 }}} \right)\frac{{\varepsilon _{\text{t}} }}{{1 - \varepsilon _{\text{t}} }},$$ where V 0 is the dead volume of the column; ε t is the total porosity of the column. The adsorption isotherms obtained are as follows: q A =1.42 c A and q B =0.19 c B , where q is concentration of component in the solid phase and c is concentration of component in the liquid phase; subscripts A and B denote PC and other components respectively. SMB experimental runs 2 Five series of experimental runs were performed with operating conditions in terms of flow rate ( Q j ), switch time ( t s ), as reported in Table 2 2 , together with the measured separation performance in terms of product purity, recovery, desorbent consumption and productivity. Note: Feed concentration of series A, B, C, D is C 0 (9.0 g/L), and that of series E is 1.5 C 0 (13.5 g/L); Q F is the feed flow rate; P R and P E are the purities of the raffinate and the extract 2 Run 5 Flow rates (ml/min) 2 t s (min) 2 Purity (%) 2 Recovery (%) 2 Desorbent consumption (L/g) 2 Productivity (g/(h·L)) Q 1 Q 2 Q 3 Q 4 Q F P R P E A1 6.5 3.6 4.1 3.1 0.5 6 98.0 71.9 96.3 0.782 1.039 A2 6.5 3.9 4.4 3.1 0.5 6 97.2 73.4 96.2 0.783 1.037 A3 6.5 4.2 4.7 3.1 0.5 6 96.9 81.5 95.3 0.791 1.027 A4 6.5 4.5 5.0 3.1 0.5 6 95.1 89.4 93.5 0.806 1.008 A5 6.5 4.8 5.3 3.1 0.5 6 91.8 90.2 85.5 0.881 0.922 B1 6.5 4.5 5.0 3.1 0.5 4 70.1 50.2 47.4 1.589 0.511 B2 6.5 4.5 5.0 3.1 0.5 5 92.6 81.8 88.2 0.854 0.951 B3 6.5 4.5 5.0 3.1 0.5 6 95.1 89.4 93.5 0.806 1.008 B4 6.5 4.5 5.0 3.1 0.5 8 93.5 83.6 89.5 0.842 0.965 B5 6.5 4.5 5.0 3.1 0.5 10 90.0 75.7 84.2 0.895 0.908 C1 6.8 4.5 5.5 3.4 1.0 4 66.5 40.8 54.0 0.697 1.165 C2 6.8 4.5 5.5 3.4 1.0 5 91.1 78.3 85.3 0.441 1.841 C3 6.8 4.5 5.5 3.4 1.0 6 94.0 83.0 90.2 0.418 1.945 C4 6.8 4.5 5.5 3.4 1.0 8 89.8 75.2 83.8 0.449 1.808 C5 6.8 4.5 5.5 3.4 1.0 10 88.1 67.4 82.3 0.458 1.775 D1 7.0 4.5 6.0 3.6 1.5 4 65.3 39.3 53.0 0.474 1.714 D2 7.0 4.5 6.0 3.6 1.5 5 82.7 71.1 70.3 0.357 2.276 D3 7.0 4.5 6.0 3.6 1.5 6 88.7 75.6 81.8 0.307 2.646 D4 7.0 4.5 6.0 3.6 1.5 8 84.5 72.8 74.4 0.337 2.407 D5 7.0 4.5 6.0 3.6 1.5 10 78.5 68.0 61.3 0.410 1.983 E1 6.8 4.5 5.5 3.4 1.0 4 64.8 38.0 51.0 0.354 2.297 E2 6.8 4.5 5.5 3.4 1.0 5 89.9 68.4 85.6 0.293 2.769 E3 6.8 4.5 5.5 3.4 1.0 6 90.6 78.5 84.7 0.296 2.742 E4 6.8 4.5 5.5 3.4 1.0 8 87.2 73.5 78.6 0.319 2.544 E5 6.8 4.5 5.5 3.4 1.0 10 85.3 65.2 78.1 0.321 2.528 RESULTS AND DISCUSSION 1 The following criteria were used to evaluate the separation performance: purity of PC in the extract $$100C_{\text{E}}^A /(C_{\text{E}}^A {\text{ + }}C_{\text{E}}^B )$$ ; recovery of PC $$100Q_{\text{E}} C_{\text{E}}^A /Q_{\text{F}} C_{\text{F}}^A$$ ; desorbent consumption $$Q_{\text{D}} /C_{\text{E}}^A Q_{\text{E}}$$ ; and productivity of stationary phase $$C_{\text{E}}^A Q_{\text{E}} /V_{\text{S}}$$ . These criteria were affected by flow rate in each section, switching time, feed flow rate and feed concentration. Five series of experiments were carried out. In each series only one factor was changed to study its effect on the performance. Effect of flow rate of Q 2 and Q 3 2 The effect of flow rate of Sections 2 ( Q 2 ) and 3 ( Q 3 ) are listed in series A (Table 2 2 ), and shown in Fig. 2 2 . With other flow rates constant, Q 2 and Q 3 related to each other. In other words, only one of these two flow rates could change independently. The results of series A showed that PC purity increases from 71.9% to 90.2% as Q 2 (and Q 3 ) increases. Since the feed flow rate was constant, increasing Q 2 (and Q 3 ) implied that the desorbent flow rate was increasing. In consequence, the concentration front of PC and other lipids would move toward the raffinate outlet. As a result, the PC concentration in both the extract and the raffinate became higher, i.e., PC purity became higher and the “purity” of other lipids lower (from 98.0% to 91.8%). On the other hand, since more PC moved into the raffinate, PC recovery became lower (from 96.3% to 85.5%). The productivity based on the feed should not change with the flow rate in the bed, but the productivity based on produced PC changed with recovery by the same ratio, i.e., decreased from 1.039 g/(h·L) to 0.922 g/(h·L). The desorbent consumption based on unit weight of PC in the extract became greater (from 0.782 L/g to 0.881 L/g) because the recovery of PC decreased. Effect of switching time 2 The effect of switching time are listed in series B, C, and D (Table 2 2 ), and shown in Fig. 3 3 showing a peak in the curve of purity versus switching time t s corresponding to purity peak was 6 min, and almost did not change with either the flow rate or the feed concentration. In the case of t s greater than 6 min, and as the solvent in Section 4 between the raffinate and the desorbent port had not been cleaned before the solvent was recycled to Section 1, the extract was contaminated by the other components. This should result in worse separation so that the purity decreased. And at t s down to 4 min, the purity of both the extract and the raffinate decreased drastically too. This could be explained by the fact that stationary phase in Section 1 between the desorbent and the extract port had not been fully desorbed before it was shifted to Section 4 so the raffinate was contaminated by PC. The relation of the recovery and productivity to t s had the same trend as purity. In the meantime, the relation of desorbent consumption to t s had reverse trend, i.e. had a minimum. Effect of feed flow rate 2 The effect of the feed flow rate can be seen in Fig. 3 3 too. In all these runs the desorbent flow rate remained at the same value of 3.4 ml/min, and Q 2 at 4.5 ml/min. As feed flow rate increased, Q 3 increased with the same increment. As a result, the separation in Section 3 became worse. The results showed that increasing feed flow increased productivity but that the purity and recovery of the extract (PC) and raffinate decreased. Effect of feed concentration 2 The effect of feed concentration can be seen by comparing the results of series C and E (Table 2 2 ), and shown in Fig. 4 4 . As the feed concentration increased, the crude phospholipids loading increased, which caused purities and PC recovery in the extract to get worse while productivity increased. According to the triangle method (Storti et al., 1993 1993 ; Mazzotti et al., 1997 1997 ; Migliorini et al., 1998 1998 ), for complete separation in SMB, the following conditions should be fulfilled: Section 1: $$m_1 \geqslant {{q_A^1 } \mathord{\left/{\vphantom {{q_A^1 } {c_A^1 }}} \right.\kern-\nulldelimiterspace} {c_A^1 }}$$ ; $$m_1 \geqslant {{q_B^1 } \mathord{\left/{\vphantom {{q_B^1 } {c_B^1 }}} \right.\kern-\nulldelimiterspace} {c_B^1 }}$$ , Section 2: $${{q_B^2 } \mathord{\left/{\vphantom {{q_B^2 } {c_B^2 }}} \right.\kern-\nulldelimiterspace} {c_B^2 }} \leqslant m_2 \leqslant {{q_A^2 } \mathord{\left/{\vphantom {{q_A^2 } {c_A^2 }}} \right.\kern-\nulldelimiterspace} {c_A^2 }}$$ , Section 3: $${{q_B^3 } \mathord{\left/{\vphantom {{q_B^3 } {c_B^3 }}} \right.\kern-\nulldelimiterspace} {c_B^3 }} \leqslant m_3 \leqslant {{q_A^3 } \mathord{\left/{\vphantom {{q_A^3 } {c_A^3 }}} \right.\kern-\nulldelimiterspace} {c_A^3 }}$$ , Section 4: $$m_4 \leqslant {{q_A^4 } \mathord{\left/{\vphantom {{q_A^4 } {c_A^4 }}} \right.\kern-\nulldelimiterspace} {c_A^4 }}$$ ; $$m_4 \leqslant {{q_B^4 } \mathord{\left/{\vphantom {{q_B^4 } {c_B^4 }}} \right.\kern-\nulldelimiterspace} {c_B^4 }}$$ , where $$m_j {\text{ = }}\frac{{{\text{net fluid flow rate}}}}{{{\text{adsorbed - phase flow rate}}}}{\text{ = }}\frac{{Q_j t_{\text{s}} - V\varepsilon _{\text{t}} }}{{V(1 - \varepsilon _{\text{t}} )}}$$ . The ideal complete separation region in the m 2 ~ m 3 plane (solid triangle) determined by triangle method and the experimental plots are shown in Fig. 5 5 showing that the plot of best separation (A4) locates at m 2 =0.67 and m 3 =0.89. It is in the vicinity of the triangle base middle where the performances of the plots of other conditions get worse depart from the best condition. This fact shows that the triangle method is useful for getting initial guesses of the optimal operating conditions. On the other hand, the performances of the conditions within the triangle other than A4 are not as satisfactory as expected by the method because the method is based on the equilibrium theory and the mass-transfer effect was not considered. Mass transfer resistance would lower the column efficiency so that the range of optimal operating conditions would be much less than the ideal one. □ represents series A; + represents series B; Δ represents series C and E; ○ represents series D; For each series, from lower-left to upper-right, represents 1 to 5 respectively. The solid line indicates the linear complete separation region according to triangle method CONCLUSION 1 High-purity PC was obtained in the extract of SMB. The separation performance of the unit and the effect of operating parameters, such as flow rates, switching time, feed flow rate and feed concentration, were studied under various operating conditions. Regions leading to complete separation were observed and explained theoretically. And the triangle theory only gives initial guesses for the optimal operating conditions. References JV Amari , PR Brown . Isolation and purification of lecithin by preparative high-performance liquid chromatography . Journal of Chromatography A , 1990 . 517 ( 1 ): 219 - 228 . DOI: 10.1016/S0021-9673(01)95723-1 http://doi.org/10.1016/S0021-9673(01)95723-1 . Isolation and purification of lecithin by preparative high-performance liquid chromatography P Bruce . Simulated moving bed processing: escape from the high-cost box . Journal of Chromatography A , 1998 . 827 43 - 160 . . Simulated moving bed processing: escape from the high-cost box A Jupke , A Epping , H Schmidt-Traub . Optimal design of batch and simulated moving bed chromatographic separation processes . Journal of Chromatography A , 2002 . 944 ( 1-2 ): 93 - 117 . 11831767 DOI: 10.1016/S0021-9673(01)01311-5 http://doi.org/10.1016/S0021-9673(01)01311-5 . Optimal design of batch and simulated moving bed chromatographic separation processes Y Li . A Study of Analysis and Purification of Soybean Phosphatidylcholine , 2004 . : Hangzhou Zhejiang University , Master Thesis, (in Chinese) . A Study of Analysis and Purification of Soybean Phosphatidylcholine M Mazzotti , G Storti , M Morbidelli . Optimal operation of simulated moving bed units for nonlinear chromatographic separations . Journal of Chromatography A , 1997 . 769 ( 1 ): 3 - 24 . DOI: 10.1016/S0021-9673(97)00048-4 http://doi.org/10.1016/S0021-9673(97)00048-4 . Optimal operation of simulated moving bed units for nonlinear chromatographic separations PV Meeren , J Vanderdeelen , M Huys , L Baert . Optimization of the column load ability for the preparative HPLC separation of soybean phospholipids . JAOCS , 1990 . 67 815 - 820 . . Optimization of the column load ability for the preparative HPLC separation of soybean phospholipids C Migliorini , M Mazzotti , M Morbidelli . Continuous chromatographic separation through simulated moving beds under linear and nonlinear conditions . Journal of Chromatography A , 1998 . 827 ( 2 ): 161 - 173 . DOI: 10.1016/S0021-9673(98)00643-8 http://doi.org/10.1016/S0021-9673(98)00643-8 . Continuous chromatographic separation through simulated moving beds under linear and nonlinear conditions DM Ruthven . Principles of Adsorption and Adsorption Processes , 1984 . : New York Wiley , 380 - 409 . . Principles of Adsorption and Adsorption Processes DM Ruthven , CB Ching . Counter-current and simulated counter-current adsorption separation process . Chemical Engineering Science , 1989 . 44 ( 5 ): 1011 - 1038 . DOI: 10.1016/0009-2509(89)87002-2 http://doi.org/10.1016/0009-2509(89)87002-2 . Counter-current and simulated counter-current adsorption separation process G Storti , M Mazzotti , M Morbidelli , S Carra . Robust design of binary countercurrent adsorption separation processes . AICHE J , 1993 . 39 ( 3 ): 471 - 492 . DOI: 10.1002/aic.690390310 http://doi.org/10.1002/aic.690390310 . Robust design of binary countercurrent adsorption separation processes 10.1631/jzus.2006.B0559 Project (No. 20040335045) supported by the Specialized Research Fund for the Doctoral Program of Higher Education of China 2005-09-27 2005-12-10 2006-06-14 2022-01-05 Copyright © 2006, Journal of Zhejiang University Science 2006 Journal of Zhejiang University-SCIENCE B (Biomedicine and Biotechnology) 07 07