Breakthrough Curve Performance Using Plate-and-Frame Affinity

Two operating modes were adopted in the elution process: cross-flow and dead-end (the retentate was closed). Moreover, different inlet flow rates were...
0 downloads 0 Views 101KB Size
854

Ind. Eng. Chem. Res. 2001, 40, 854-861

Breakthrough Curve Performance Using Plate-and-Frame Affinity-Membrane Modules Yi-Da Tsai and Shing-Yi Suen* Department of Chemical Engineering, National Chung Hsing University, Taichung 402, Taiwan

The effects of inlet flow rate on the breakthrough curve performance of lysozyme using plateand-frame dye-affinity-membrane modules were investigated in this work. For the nonadsorption results, it is found that the influences of flow rate on both permeate and retentate breakthrough curves mainly came from the flow maldistribution and extra-module mixing effects. On the other hand, the effects of axial and radial diffusion were negligible in this case. When single-protein adsorption occurred, the elution peak height in the permeate decreased with increasing flow rate, which resulted from the limitation of convective flow. Moreover, this study investigated three larger-scale designs: a multiple-membrane stack in one module, two modules connected in parallel, and two modules in tandem. Both of the two-module designs resulted in a greater amount of eluted protein, which is attributed to their lower flow rates in the modules than the multiple-membrane-stack design. As for the two-protein breakthrough curve performance, a simple separation of lysozyme from nonadsorbable bovine serum albumin (BSA) was observed but the existence of BSA declined the lysozyme adsorption. The degree of influence on lysozyme adsorption was the same in both one-module and two-module-in-tandem designs. Furthermore, the recovered percentages of lysozyme in both one-module and two-module designs were low, and this was mainly restricted by the limitation of flow rate for the membrane modules adopted in this study. 1. Introduction Applying adsorptive membrane technique as a practical and useful bioseparation method has gained much attention because of its lower mass-transfer limitations than the traditional column technique.1-4 Different shapes of adsorptive membranes such as membrane disk, hollow fiber, or spiral-wound membrane and their effects on adsorption and separation have been extensively investigated.3,4 Among them, membrane disks and hollow fibers are most frequently utilized and studied in the literature.5-20 The major advantages of membrane disks include various membrane products in the market, uncomplicated surface modification methods, and a simpler and cheaper design of the disk holder, as compared to the other types of membranes. However, the practical applications of membrane disks are usually restricted by their inability to deal with crude solutions, the problems of flow maldistribution, and disk edge leaking in larger-scale separations.3 On the other hand, hollow fibers are considered better adsorptive supports because of their high specific surface area leading to a higher relative adsorption capacity. In addition, the cross-flow cartridge design of hollow fibers is feasible to induce an effective separation for crude solutions. The development of adsorptive hollow fibers, however, may suffer the problem of nonuniform ligand distribution if the ligand is immobilized onto the fibers entrapped in a cartridge.16 Besides, it can also lead to a difficulty in assembling the fibers into a cartridge after a uniform ligand immobilization onto individual fibers. Consequently, this emerging separation technique still re* To whom correspondence should be addressed. Tel: 8864-2852590. Fax: 886-4-2854734. E-mail: [email protected].

quires further improvement with proper design and operation of its corresponding adsorptive-membrane systems. A cross-flow design of flat-sheet membranes would combine the advantages previously described for both flat-sheet membranes and hollow fibers and accordingly may achieve a more efficient separation performance. In other words, a plate-and-frame membrane module design can be advantageously suggested for adsorption. Therefore, this work aims to investigate the breakthrough curve performance using plate-and-frame affinity-membrane modules. Two proteins, lysozyme and bovine serum albumin (BSA), were adopted, and the immobilized ligand was Cibacron Blue 3GA. System parameters of interest in this study include batch adsorption properties, intrinsic adsorption rate, pressure, flow rate, axial diffusion, radial diffusion, and extra-module effects. Moreover, different larger-scale designs were further tested for an extensive study on the performance-influencing factors. 2. Experimental Section Materials. The Immobilon AV (IAV) membrane of Millipore Co. (Bedford, MA), with an average pore size of 0.65 µm and a thickness of 140 µm, was adopted in this study. Cibacron Blue 3GA (C9534, MW 774.2), chicken egg white lysozyme (L6876, MW 14 300), and BSA (A3059, MW 67 000) were purchased from Sigma Chemical (St. Louis, MO). 1,10-Diaminodecane, as a spacer arm for ligand immobilization, was obtained from TCI Co. (Tokyo, Japan). Spacer Arm Conjugation and Ligand Immobilization. The IAV membrane has been chemically activated for immobilizing ligands with amine groups. Upon lack of active amine groups on Cibacron Blue 3GA, a

10.1021/ie0005912 CCC: $20.00 © 2001 American Chemical Society Published on Web 01/05/2001

Ind. Eng. Chem. Res., Vol. 40, No. 3, 2001 855

Figure 1. Schematic diagram for a plate-and-frame membrane module used in this work.

straight-chain diamine molecule was employed on this membrane as a spacer arm for a linkage binding. 1,10Diaminodecane was chosen in this work because of its good performance on the degree of Cibacron Blue 3GA immobilization and the relative adsorption capacity of lysozyme, as indicated in a previous paper by the authors.21 Prior to the reactions, the IAV membrane was cut into 9 cm × 9 cm pieces. 1,10-Diaminodecane (0.1 mol) was dissolved into 60 mL of 0.5 M phosphate buffer (pH 7.2) and reacted with one piece of dry IAV membrane. The reaction was then conducted at room temperature for 12 h. After reaction, the membrane was thoroughly rinsed to remove the unreacted diamine. The spacer-bound membrane was then incubated with 60 mL of a 1.65% (w/v) Cibacron Blue 3GA solution under 60 °C. After 30 min, 60 mL of a 6% (w/v) NaCl solution was added into the mixture and the reaction was continuously conducted at the same temperature for 1 h. A total of 60 mL of a 2% (w/v) Na2CO3 solution was finally added, and the temperature was raised to 80 °C for another 1 h incubation. After reaction, the membrane was washed with deionized water until no blue dye was eluted. Breakthrough Curve Experiments. Figure 1 illustrates the acrylic plate-and-frame module used in this study. The size of the square module is 14.6 cm × 14.6 cm, and the membrane size matching the module is 9 cm × 9 cm. A square ring of silicone was used for tighter compression of the module to prevent fluid from laterally leaking. Other equipment for breakthrough curve experiments included a pump consisting of a motor drive (MasterFlex L/S 07520-00, Cole-Parmer,

Barrington, IL) with a pump head (MasterFlex L/S 07016-20), two absorbance detectors (UA-6, ISCO, Lincoln, NE, and AC-5200L, ATTO, Tokyo, Japan), and two dataloggers (MINILOG, 3D, London, England, and ESCORT, Cox Technologies, Belmont, NC). The protein absorbance was detected at 280 nm. Two fraction collectors (Retriever 500, ISCO) were used to collect effluent samples for the two-protein separations. Moreover, two pressure gauges were employed for measuring the pressures in the inlet and retentate. Figure 2 schematically shows the diagram of the breakthrough curve experiment design. All breakthrough curve experiments were conducted at room temperature, and each experiment was repeated twice, except for the twoprotein ones. The buffer used for loading and washing in adsorption experiments was 50 mM Tris-HCl, pH 7, with 0.005% NaN3. The elution buffer was the loading buffer with 1 M KCl. Both buffers were vacuum-filtered with 0.2 µm nylon membranes (Lida Manufacturing, Kenosha, WI). Protein solutions were prepared with the above buffers and filtered by 0.45 µm filters (Millex-HV, Millipore, Bedford, MA). All solutions were degassed prior to use. (a) Water Permeability Measurements. A Cibacron Blue 3GA-immobilized membrane was placed in a plate-and-frame module, and the permeate was opened to the atmosphere. With variation of the inlet flow rate of deionized water, the flow rates in the two outlets and the pressures in both the inlet and retentate were recorded. (b) Nonadsorption Experiments. Nonadsorption experiments were conducted using 0.2 mg/mL lysozyme in a pure elution buffer. After the module with an affinity membrane was equilibrated with an elution buffer, the protein solution was loaded and the absorbances in both outlets were recorded by dataloggers. When the raised absorbance from the permeate was steady, the elution buffer was loaded as a washing buffer to bring the absorbance curve back to the baseline. The effect of variation of the inlet flow rate on the breakthrough curve was then investigated. (c) Single-Protein Adsorption Experiments. In single-protein adsorption experiments, the membrane module was equilibrated with the loading buffer, followed by loading of the 0.2 mg/mL lysozyme solution prepared in the loading buffer. The protein absorbances in both outlets were detected and also recorded by dataloggers. After the absorbance from the permeate did not rise, the loading buffer was pumped through as a washing buffer until the absorbance returned to the baseline. The elution buffer was then used to elute the bound protein from the system, and this elution step was finally stopped when the absorbance from the permeate returned to the baseline again. Two operating modes were adopted in the elution process: cross-flow and dead-end (the retentate was closed). Moreover, different inlet flow rates were used to evaluate the flow rate effect. For a larger-scale performance, three different designs were investigated. In the first design, multiple dyeaffinity membranes as a stack were inserted into one membrane module. In the other two designs, two modules with one affinity membrane in each module were connected in different ways: parallel and tandem. The schematic diagrams were plotted in Figure 2. The experimental procedures for larger-scale designs were the same as those previously described. The inlet flow

856

Ind. Eng. Chem. Res., Vol. 40, No. 3, 2001

Figure 2. Schematic diagram for breakthrough curve experiments using plate-and-frame membrane module and two-module-in-connection designs.

rate was 13 mL/min, and the dead-end mode was adopted in the elution process. (d) Two-Protein Adsorption Experiments. In the two-protein experiments, the inlet concentration was 0.2 mg/mL for lysozyme and 0.2 mg/mL for BSA. The inlet flow rate was 13 mL/min, and the dead-end mode was adopted in the elution process. Using the same experimental procedures as those in single-protein experiments, the effluent solutions were collected by fraction collectors. To avoid detection difficulty in low protein concentrations, the collected effluents were concentrated by a rotary evaporator prior to FPLC analysis under 280 nm. The FPLC system (Amersham Pharmacia Biotech AB, Uppsala, Sweden) was composed of a controller (LCC-501), a pump (P-500), a UV detector (UV-1), and a gel filtration column (Superose 6PC 3.2/ 30). The mobile phase was 0.1 M Tris-HCl, pH 7.4, with 0.02% NaN3, and the flow rate was 0.42 mL/min. The sample amount for each injection was 20 µL. As for the larger-scale two-protein separation, the design of two modules connected in tandem was adopted. 3. Results and Discussion Batch Adsorption Performance. In this study, 1,10-diaminodecane was used as a spacer arm for the immobilization of Cibacron Blue 3GA onto the IAV membranes. This immobilization is primarily based on the binding of the triazine group of Cibacron Blue 3GA with the free amine group of the spacer molecule. The uniform intense blue displayed on the outer membrane surface proved the dye-ligand immobilization to be successful. Batch adsorption experiments of lysozyme and BSA onto dye-immobilized membranes were not conducted in this work. Instead, the results reported in the authors’ earlier works21,22 using the same membrane (size 1 cm × 1 cm), spacer arm, and immobilization technique would be directly applied for later analysis. It may be worth noting that, in the previous works, the adsorbed protein amount calculated from the difference

between the original and remaining protein amount in the solution was compared with the collected eluted amount using the elution buffer. Both amounts were in close agreement. This consistency implies that the elution buffer could completely elute the adsorbed protein out of the affinity membranes. In the previous studies, the Langmuir model was employed to fit the data of the isotherm and association curve. The fitted parameter values for lysozyme were saturation capacity (qm) ) 0.65 ( 0.04 mg/cm2 (based on the frontal area of the membrane), dissociation equilibrium constant (Kd) ) 0.06 ( 0.02 mg/mL, and association rate constant (ka) ) 0.034 min-1 (mg/mL)-1. From the saturation capacity data, we can estimate the saturation capacity of a dye-affinity membrane with 81 cm2 of frontal area as 52.65 mg of lysozyme. Moreover, the nonspecific binding of the lysozyme was measured and found to be negligible in the previous works. As to the association rate constant, the data is at the same order of magnitude as that for higher dye-ligand density using gel beads as solid supports in the literature.25 As discussed in the literature, the rate constant value increased with decreasing ligand density. Therefore, it may be concluded that the slow kinetics for lysozyme binding resulted from the steric hindrance of the crowded ligand sites on the dye-ligand membranes. On the other hand, the amount of adsorbed BSA on the Cibacron Blue 3GA-immobilized membrane was negligible, which possibly resulted from the effect of steric hindrance. Effects of Inlet Flow Rate on Water Permeability. Figure 3 presents the flow rate variations in the two outlets with variation of the inlet flow rate of water through a plate-and-frame membrane module. The flow rate from the retentate was 55% of the inlet, while the permeate flow rate was 44%. The overall outlet flow rate, retentate plus permeate, was almost identical with the inlet rate. These observations demonstrate that no liquid was lost during the flow operations and the flow rate ratio of permeate to retentate

Ind. Eng. Chem. Res., Vol. 40, No. 3, 2001 857

Figure 3. Relationship between the inlet and outlet flow rates using a plate-and-frame membrane module.

was linear, independent of the inlet flow rate. In this study, the highest inlet flow rate available was 106 mL/ min; beyond that lateral leaking was observed. On the other hand, the lowest flow rate was 13 mL/min, below which liquid could hardly pass through the permeate because of the limitations from the apparatus and module design. The limitations came from the difference between the resistances intrinsically existing in the retentate and permeate sides. The retentate outlet was open at the same side as that of the inlet, whereas the permeate was at the other side (as shown in Figure 1). If the flow rate is not great enough to overcome the resistance difference and to create a positive transmembrane pressure, the liquid will flow out of the lower resistance retentate outlet. This is why the liquid could only be eluted from the permeate until the flow rates were above 13 mL/min. The effect of the inlet flow rate on the transmembrane pressure was also investigated in this study. The readings from both pressure gauges in the inlet and retentate were less than 0.07 MPa (1 psi), even under the highest inlet flow rate operation. Moreover, the pressures remained negligible in either the dead-end mode or larger-scale designs. Consequently, transmembrane pressures were insignificant in this plate-andframe module using microfiltration membranes. Nonadsorption Breakthrough Curves. In the nonadsorption experiments, 0.2 mg/mL lysozyme in an elution buffer was loaded and then emerged into the effluents without adsorption. The results of different inlet flow rates are presented in Figure 4. The flow rate did not significantly affect the breakthrough curve from the permeate, with only the effluent volume to reach c (effluent concentration)/c0 (inlet concentration) ) 1 being altered slightly. The measured effluent volume to reach c/c0 ) 1 was the largest for a flow rate of 13 mL/min, indicating that the broadest breakthrough curve occurred at the lowest flow rate. As for the retentate, the flow rate effect on the nonadsorption breakthrough curve was more significant and c/c0 ) 1 can only be reached at high flow rates. It may be noted that the longer delay volume in the breakthrough curve for the retentate comes from its greater flow rate than that of the permeate. To explain the flow rate effect, the possible masstransfer effects in a cross-flow membrane process need to be evaluated. The mass-transfer effects contributing to nonadsorption breakthrough curves can be divided into extra-module and inside-module effects.23,24 Extramodule effects include the delay volume effect from the

Figure 4. Breakthrough curves for nonadsorption experiments using one membrane module at various inlet flow rates: (a) loading; (b) washing.

fluid flowing through the tubings, pressure gauges and detector flow cells, and the mixing effect inside these accessories.14,23,24 In this work, the extra-module apparatus was kept unchanged in all of the experiments such that the extra-module delay volume should be identical for all of the cases. However, the mixing effect through these accessories may not be neglected. If a CSTR mixing model could express this mixing effect as shown in the literature,14,23 the nonadsorption breakthrough curve for the lowest flow rate would rise to c/c0 ) 1 after all of the other curves did. This interpretation seems to properly explain the nonadsorption results in this work. Convection and axial diffusion are the possible masstransfer mechanisms when the fluid flows parallelly to the membrane and emerges to the retentate. On the other hand, convection, axial diffusion, and radial diffusion inside the pores are the main mass-transfer factors for the fluid flowing perpendicularly to the membrane and emerging to the permeate. These insidemodule mass-transfer factors should be analyzed to investigate their effects. Using the membrane size (L) of 9 cm and the measured cross-sectional area perpendicular to the retentate flow (A0) of 0.08 cm2, the time scale for convection parallel to the membrane, Tconvection,parallel ) L/vretentate (v ) interstitial flow velocity) ) A0L/(A0vretentate) ) A0L/Qretentate (Q ) flow rate), was calculated as 0.1, 0.05, 0.03, 0.02, and 0.01 min for inlet flow rates from 13 to 106 mL/min, respectively. The time scale for the axial diffusion in the same flow direction, Taxial diffusion,parallel ) L2/D, was about 33 170 h by adopting the diffusion coefficient of 40.7 × 10-6 cm2/

858

Ind. Eng. Chem. Res., Vol. 40, No. 3, 2001

min for lysozyme from the literature.25 The characteristic time of the axial diffusion was much larger than the convection time, indicating that the axial diffusion for flows parallel to the membrane was insignificant and the influence of the flow rate on the retentate breakthrough curve did not come from the axial diffusion. Instead, the flow rate effect may be attributed to the phenomenon of flow maldistribution. As shown in Figure 1, the fluid inlets the membrane with a width of 9 cm from an entrance having a much smaller diameter. Therefore, the time for the fluid to fill the whole membrane is increased with decreasing flow rate. Accordingly, the resulting breakthrough curve at lower flow rate should be broader. In addition, c/c0 was not raised to 1 for lower flow rates as shown in Figure 4, which demonstrates the flow maldistribution effect. Inside the membrane pores, the time scales for convection, axial diffusion, and radial diffusion can be estimated as Tconvection,pore ) Lm/vpermeate ) ApLm/ Apvpermeate ) ApLm/Qpermeate (Ap ) L2), Taxial diffusion,pore ) Lm2/D, and Tradial diffusion,pore ) Rm2/D, respectively. Using the values of membrane thickness (Lm) ) 140 µm, porosity () ) 0.7, and average pore radius (Rm) ) 0.325 µm for the IAV membrane, the estimated time scales were Taxial diffusion,pore ) 4.8 min, Tradial diffusion,pore ) 2.6 × 10-5 min, and Tconvection,pore ) 0.14-0.02 min in the experimental flow-rate range. The radial diffusion time was much smaller than the convection and axial diffusion times, indicating that the radial diffusion in the membrane pores was very fast and did not influence the mass-transfer rate. Moreover, the axial diffusion time scale was 30-200 times greater than the convection time, meaning that axial diffusion of protein took much longer time than flowing through the pores so that the axial diffusion effect was insignificant. Similar to the retentate breakthrough curve, the flow-rate effect on the permeate breakthrough curve may also result from flow maldistribution across the wide membrane, but the effect was less important. Single-Protein Breakthrough Curves. (a) Different Operating Modes. In the single-protein adsorption experiments, part of the adsorbed protein molecules on the frontal area of the membrane were eluted from the retentate if the cross-flow mode was adopted in the elution stage. In this case, the eluant may be diluted because of its division into two outlet streams. To prevent protein elution from the retentate, two methods were applied in this study. The first method is placing blank IAV membranes before the affinity membrane. Using this method, the protein amount eluted from the retentate could be reduced, but at the same time, the eluted amount from the permeate was also decreased because of the more significant effect of axial diffusion (the results are not shown). Moreover, the more blank membranes were used, the lower protein amounts were eluted from both retentate and permeate. The second method is to seal the retentate while eluting, i.e., employing the dead-end mode. In this mode, all of the eluted protein molecules can only flow out from the permeate. Because of the fact that the first method could not efficiently terminate protein elution from the retentate, the dead-end operation is considered to be a better way in application and was adopted in the subsequent flow experiments. (b) Effects of Flow Rate. Figure 5 shows the singleprotein breakthrough curves for 0.2 mg/mL lysozyme in loading buffer at various flow rates with the use of

Figure 5. Breakthrough curves for single-protein experiments using one membrane module at various inlet flow rates: (a) loading; (b) washing; (c) eluting.

the dead-end mode in elution. The highest inlet flow rate was taken as 80 mL/min, instead of 106 mL/min as before, because the flow rate beyond this value would cause lateral leaking under the dead-end operation mode. In Figure 5, the retentate breakthrough curves and the loading parts of the permeate ones were not as broad as the nonadsorption curves in Figure 4, and the broadness for different flow rates was in the opposite order. This indicates that the effects of extra-module mixing and flow maldistribution became less important when adsorption took place. To theoretically investigate the competition between convective flow and the intrinsic adsorption rate, the characteristic time of adsorption, Tadsorption ) 1/kac0, was calculated using the association rate constant obtained previously. The evaluated adsorption time was 147.1 min, which is much longer than the convection times for both outlets. Accordingly, the protein molecules had no sufficient time to adsorb before they flew through the outlets even under the lowest feasible flow rate. With the flow rate increasing, protein got less time to adsorb, which resulted in a broader breakthrough curve. Consequently, convective flow dominated the whole adsorption process. In Figure 5, the elution peak became broader and the height decreased with an increase in the flow rate. Similar to the loading stage, convective flow dominated the elution process. To more precisely investigate the flow-rate effect, the area under the elution peak was integrated and the eluted protein amount was evaluated. The results are listed in Table 1. The eluted protein amount was reduced with an increase in the flow rate. When the elution is complete, the eluted protein amount is equal to the desorbed amount and so is the

Ind. Eng. Chem. Res., Vol. 40, No. 3, 2001 859 Table 1. Effect of Flow Rate on the Eluted Protein Amount in Single-Protein Performance permeate flow rate in cross-flow mode permeate flow rate eluted (loading and in dead-end mode lysozyme inlet flow rate washing stages) (elution stage) amount (mL/min) (mL/min) (mL/min) (mg) 13 27 40 57 80

6 11.5 18 25 36

13 27 40 57 80

0.94 0.62 0.46 0.38 0.32

Table 2. Eluted Protein Amounts Using the Membrane Stack in One Membrane Module permeate flow rate permeate in cross-flow flow rate mode in dead-end eluted inlet (loading and mode lysozyme flow rate washing stages) (elution stage) no. of amount (mL/min) (mL/min) (mL/min) membranes (mg) 13

6

13

27

11.5

27

80

36

80

1 2 3 1 2 3 1 2 3

0.94 1.08 0.84 0.62 0.80 0.62 0.32 0.36 0.36

Table 3. Eluted Protein Amounts Using Different Connection Designs with Two Membrane Modules at 13 mL/min

connection design

permeate flow rate in cross-flow mode (loading and washing stages) (mL/min)

permeate flow rate in dead-end mode (elution stage) (mL/min)

eluted lysozyme amount (mg)

parallel tandem

6 5.5

13 13

1.56 1.60

adsorbed amount. For higher flow rate, less protein could be adsorbed because of the limitation of the convective flow and a smaller protein amount was eluted. In this work, the eluted protein amount further represents the dynamic capacity for the flowing adsorption process. The eluted amounts or dynamic capacities presented in Table 1 were much lower than the static capacity (40.5 mg) of the affinity membrane at an equilibrium concentration of 0.2 mg/mL determined by the Langmuir isotherm model. (c) Different Larger-Scale Designs. Three largerscale designs were investigated in this study: stacking two or three affinity membranes in one membrane module, connecting two modules in parallel with one affinity membrane in each module, and connecting two modules in tandem. The results of the eluted protein amount are listed in Tables2 and 3. In Table 2, stacking more membranes did not significantly raise the eluted protein amount or, in other words, the adsorbed protein amount. As discussed above, the flow rates used in this study were too fast to permit sufficient adsorption. The residence time raised by stacking more affinity membranes in the module was negligible and still not sufficient for more protein molecules to adsorb. Therefore, the adsorbed protein amount was essentially not increased with the number of stacked membranes. The eluted protein amounts displayed in Table 3 for both two-module-in-connection designs were greater

Figure 6. Two-protein breakthrough curves using one membrane module at 13 mL/min: (a) loading; (b) washing; (c) eluting.

than the result in Table 2 for the two-membrane stack in one module at a flow rate of 13 mL/min. The better performance may be attributed to the lower flow rate through the membranes for these module-in-connection designs. In the parallel design, the individual flow rate getting into each membrane module was 6.5 mL/min and the flow rate from the permeate of each module was around 3 mL/min. In the tandem design, the flow rate entering the first module was 13 mL/min and the flow rate out of the permeate and inleting the second module was 6.5 mL/min (by subtracting the measured retentate-1 flow rate from the inlet flow rate). The measured permeate flow rate from the second module was 5.5 mL/ min. These flow rates were lower than the inlet flow rate (13 mL/min) and the permeate flow rate (6 mL/ min) for the design of stacking two membranes in one module. Accordingly, the protein molecules got a longer time to adsorb in the module-in-connection designs. More specifically, the tandem design led to a slightly greater amount of eluted protein than the parallel design. Although the flow rates through the membranes were higher for the tandem design, its longer distance from the inlet to the permeate of the second module had prolonged the retention and adsorption time and eventually compensated the flow-rate effect to surpass the parallel design in the adsorbed protein amount. Two-Protein Breakthrough Curves. Two-protein breakthrough curves for a feed containing 0.2 mg/mL lysozyme and 0.2 mg/mL BSA using one membrane module at 13 mL/min are depicted in Figure 6. The twoprotein solution was essentially a ternary solute solution because approximately 15% of the BSA existed as a dimer and the rest as a monomer. It is worth recalling that lysozyme could be bound onto the dye-affinity membrane but BSA adsorption was insignificant. There-

860

Ind. Eng. Chem. Res., Vol. 40, No. 3, 2001

fore, it can be expected that a simple separation of lysozyme from BSA would occur during this operation. The resulting breakthrough curves exactly demonstrate this phenomenon. For both of the permeate and retentate results in the loading stage, three solutes almost emerged at the same time and their effluent concentration rose to a value of around 80-95% of their feed. The washing curves were almost identical for these three solutes. The final elution peak in the permeate contained only lysozyme, confirming negligible binding of BSA. The elution peak of lysozyme from FPLC analysis was not consistent with the one recorded by the datalogger, which may result from the experimental error by concentrating the effluents prior to FPLC analysis. Consequently, the eluted lysozyme amount was evaluated by integrating the recorded peak from the datalogger instead of using the data from FPLC analysis. The value of the eluted amount was 0.78 mg, about 83% of the amount in the single-protein experiment (0.94 mg). This result implies that the existence of BSA in solution has affected the adsorbed amount of lysozyme. It is possible that, although BSA adsorption was found negligible, there may be a small BSA amount (but undetectable) weakly bound on the membrane to compete with lysozyme for adsorption. In addition, BSA molecules in solution might interact with lysozyme and influence the lysozyme adsorption. The two-protein breakthrough performance using the connection-in-tandem design is shown in Figure 7. Low flow rate (around 1 mL/min) from the retentate in the second module resulted in a small effluent volume as shown in Figure 7. The breakthrough curves are similar to those for one-module experiments, except that the effluent concentrations for loading curves of the BSA dimer decreased. Because the concentration of the BSA dimer was the lowest among these three solutes, the low c/c0 values might only be caused by difficulties in FPLC analysis. The calculated amount of eluted lysozyme from the datalogger-recorded curve was 1.36 mg, greater than the one-module result. This value was 85% of the eluted amount in the single-protein experiment using the same design (1.60 mg). Conclusively, the lysozyme adsorption was declined because of the existence of BSA and the degree of influence remained unchanged even in the larger-scale design. In this work, pure BSA could not be collected because of close emergences for different proteins in the loading and washing stages. To resolve this close-emergence problem, it is necessary to greatly decrease the flow rate. The lower the flow rate, the larger the amount of lysozyme adsorbed and the longer the time obtained for lysozyme to emerge. This will lead to a larger offset in the emergences between lysozyme and BSA, and consequently pure BSA in the effluent may be achieved in the loading stage. To investigate the process efficiency, the cycle time and the ratio of the eluted protein amount to the loaded amount need to be calculated. From Figures 6 and 7, around 150 mL of effluent volume was needed for a separation cycle of this plate-and-frame membrane process. This value was obtained by subtracting the unnecessary waiting volume from the total volume shown in the figures. Within a cycle, about 40 mL of protein solution was needed for loading onto membranes, which was similarly calculated by subtracting the unnecessary waiting volume from the actual loading

Figure 7. Two-protein breakthrough curves using two membrane modules connected in tandem at 13 mL/min: (a) loading; (b) washing; (c) eluting.

volume. Accordingly, the loaded lysozyme amount would be 8 mg, which was obtained by multiplying the loading volume by the feed concentration of 0.2 mg/mL. The corresponding adsorbed protein amounts were 0.78 and 1.36 mg for one-module and two-module-in-tandem processes, respectively. Therefore, the recovered percentages would be 9.8% for the one-module process and 17% for the two-module-in-tandem process. These values were relatively low and mainly induced by the high flow rate restriction of the membrane module used in this work. Therefore, a more efficient protein separation performance can be expected if a better design can be employed to allow lower flow rate and the use of multiple modules connected in tandem. 4. Conclusion The breakthrough curve performance using plate-andframe affinity-membrane modules was investigated in this paper. In nonadsorption experiments, different mass-transfer mechanisms were analyzed and none of the diffusion factors was found to be dominant. The flow rate effect shown on the nonadsorption breakthrough

Ind. Eng. Chem. Res., Vol. 40, No. 3, 2001 861

curves mainly came from extra-module mixing and flow maldistribution phenomena. In single-protein experiments, convective flow dominated the loading and elution stages. The eluted protein amount decreased with an increase in the flow rate. Moreover, three larger-scale designs were investigated in this study. The results showed that the design of stacking more membranes in the same module did not raise the eluted protein amount. Conversely, both two-module-in-connection designs resulted in a greater eluted protein amount, which was attributed to the lower flow rate through the membranes. In two-protein experiments, a simple separation of lysozyme from BSA was observed. However, the existence of BSA reduced the lysozyme adsorption, and this decrease remained the same in both one-module and two-module designs. In addition, the recovered percentages using the calculated values of cycle time, loaded protein amount, and adsorbed protein amount were relatively low in this study. These performance limitations mainly came from the high flow-rate restriction by the adopted membrane module. An improved design of the membrane module to allow a lower flow rate or a multiple-module design using more modules connected in tandem is required for a more effective protein separation performance. Moreover, the dead-end configuration may be a better choice to allow a lower flow rate and a more sufficient adsorption when microfiltration membranes are used and the solute accumulation is not a problem. On the other hand, if there is a biosystem with a faster binding kinetics such as protein adsorption onto ion-exchange membranes, the faster and more adequate adsorption will be expected using the plate-and-frame module. In conclusion, the investigation of the breakthrough curve performance and the subsequent suggestions in this study may provide useful guidelines for better plate-and-frame adsorptive-membrane separations. Acknowledgment This work was sponsored by the National Science Council of Republic of China (NSC 87-2214-E-005-005). Literature Cited (1) Brandt, S.; Goffe, R. A.; Kessler, S. B.; O’Connor, J. L.; Zale, S. E. Membrane-based Affinity Technology for Commercial Scale Purifications. Bio/Technology 1988, 6, 779. (2) Briefs, K.-G.; Kula, M.-R. Fast Protein Chromatography on Analytical and Preparative Scale Using Modified Microporous Membranes. Chem. Eng. Sci. 1992, 47, 141. (3) Roper, D. K.; Lightfoot, E. N. Separation of Biomolecules Using Adsorptive Membranes. J. Chromatogr. A 1995, 702, 3. (4) Charcosset, C. Purification of Proteins by Membrane Chromatography. J. Chem. Technol. Biotechnol. 1998, 71, 95. (5) Hou, K. C.; Zaniewski, R.; Roy, S. Protein A Immobilized Affinity Cartridge for Immunoglobulin Purification. Biotechnol. Appl. Biochem. 1991, 13, 257. (6) Kim, M.; Saito, K.; Furusaki, S.; Sato, T.; Sugo, T.; Ishigaki, I. Adsorption and Elution of Bovine γ-Globulin Using an Affinity Membrane Containing Hydrophobic Amino Acids as Ligands. J. Chromatogr. 1991, 585, 45.

(7) Iwata, H.; Saito, K.; Furusaki, S.; Sugo, T.; Okamoto, J. Adsorption Characteristics of an Immobilized Metal Affinity Membrane. Biotechnol. Prog. 1991, 7, 412. (8) Nachman, M.; Azad, A. R. M.; Bailon, P. Membrane-based Receptor Affinity Chromatography. J. Chromatogr. 1992, 597, 155. (9) Nachman, M. Kinetic Aspects of Membrane-based Immunoaffinity Chromatography. J. Chromatogr. 1992, 597, 167. (10) Guo, W.; Shang, Z.; Yu, Y.; Zhou, L. Membrane Affinity Chromatography of Alkaline Phosphatase. J. Chromatogr. A 1994, 685, 344. (11) Klein, E.; Eichholz, E.; Yeager, D. H. Affinity Membranes Prepared from Hydrophilic Coatings on Microporous Polysulfone Hollow Fibers. J. Membr. Sci. 1994, 90, 69. (12) Klein, E.; Eichholz, E.; Theimer, F.; Yeager, D. Chitosan Modified Sulfonated Poly(ethersulfone) as a Support for Affinity Separations. J. Membr. Sci. 1994, 95, 199. (13) Serafica, G. C.; Pimbley, J.; Belfort, G. Protein Fractionation Using Fast Flow Immobilized Metal Chelate Affinity Membranes. Biotechnol. Bioeng. 1994, 43, 21. (14) Suen, S.-Y.; Etzel, M. R. Sorption Kinetics and Breakthrough Curves for Pepsin and Chymosin Using Pepstatin A Affinity Membranes. J. Chromatogr. A 1994, 686, 179. (15) Bueno, S. M. A.; Haupt, K.; Vijayalakshmi, M. A. Separation of Immunoglobulin G from Human Serum by Pseudobioaffinity Chromatography Using Immobilized L-Histidine in Hollow Fibre Membranes. J. Chromatogr. B 1995, 667, 57. (16) Charcosset, C.; Su, Z.; Karoor, S.; Daun, G.; Colton, C. K. Protein A Immunoaffinity Hollow Fiber Membranes for Immunoglobulin G Purification: Experimental Characterization. Biotechnol. Bioeng. 1995, 48, 415. (17) Klein, E.; Yeager, D.; Seshadri, R.; Baurmeister, U. Affinity Adsorption Devices Prepared from Microporous Poly(amide) Hollow Fibers and Sheet Membranes. J. Membr. Sci. 1997, 129, 31. (18) Denizli, A.; Salih, B.; Arica, M. Y.; Kesenci, K.; Hasirci, V.; Piskin, E. Cibacron Blue F3GA-incorporated Macroporous Poly(2-hydroxyethyl methacrylate) Affinity Membranes for Heavy Metal Removal. J. Chromatogr. A 1997, 758, 217. (19) Denizli, A.; Tanyolac, D.; Salih, B.; Aydinlar, E.; Ozdural, A.; Piskin, E. Adsorption of Heavy-metal Ions on Cibacron Blue F3GA-immobilized Microporous Polyvinylbutyral-based Affinity Membranes, J. Membr. Sci. 1997, 137, 1. (20) Weissenborn, M.; Hutter, B.; Singh, M.; Beeskow, T. C.; Anspach, F. B. A Study of Combined Filtration and Adsorption on Nylon-based Dye-affinity Membranes: Separation of Recombinant L-Alanine Dehydrogenase from Crude Fermentation Broth. Biotechnol. Appl. Biochem. 1997, 25, 159. (21) Suen, S.-Y.; Tsai, Y.-D. Comparison of Ligand Density and Protein Adsorption on Dye-Affinity Membranes Using Different Spacer Arms. Sep. Sci. Technol. 2000, 35, 69. (22) Tsai, Y.-D. Design and Investigation of Bioseparation Technology Using Plate-and-Frame Affinity-Membrane Modules. M.S. Thesis, National Chung Hsing University, Taichung, Taiwan, 1999. (23) Yang, H.; Bitzer, M.; Etzel, M. R. Analysis of Protein Purification Using Ion-Exchange Membranes. Ind. Eng. Chem. Res. 1999, 38, 4044. (24) Suen, S.-Y.; Tsai, Y.-H.; Chen, R.-L. Comparison of Breakthrough Performance Using Dye-Affinity Membrane Disks and Gel Bead Columns. Sep. Sci. Technol. 2000, 35, 573. (25) Boyer, P. M.; Hsu, J. T. Effects of Ligand Concentration on Protein Adsorption in Dye-ligand Adsorbents. Chem. Eng. Sci. 1992, 47, 241.

Received for review June 20, 2000 Revised manuscript received October 30, 2000 Accepted November 6, 2000 IE0005912