Biotechnol. Prog. 1993, 9,285-290
285
Protein Purification with Novel Porous Sheets Containing Derivatized Cellulose James L. Manganaro*itand Bruce S. Goldberg: FMC Corporation, Princeton, New Jersey 08543, and FMC BioSupport Materials, Pine Brook, New Jersey 07058
Novel porous sheets containing commercially available cellulosic ion exchange media of several different functionalities in a PVC matrix have been prepared and evaluated. The advantages of these materials are resistance to alkali, clean in place procedures and low nonspecific binding. These sheets thus provide well-known and well-characterized particulate chromatographic media in a porous sheet format. The porous sheet format permits use of short, squat geometries (stacked sheets) as well as tolerance of high pressure gradients. The net result is the ability to handle much higher flow rates (10fold or greater) than achievable by comparable cellulosic packed columns of the neat particulate media. Chromatographic separation of natural egg white proteins by a linear salt gradient on a single sheet of 1.3-mm thickness was demonstrated. In addition, a very significant advantage of the porous sheet format is that it permits the fabrication of devices which are not possible with particulate media. An example of such a device is the Acti-Mod Spiral Module. In the spiral module a spiral flow channel is formed by wrapping the porous sheet material around a mandrel. Embossed ribbing in the porous sheet provides channel spacing. The small open channel accommodates high flow rates of biological suspensions while the porous sheet walls selectively remove components of the suspension.
Introduction The growing number and ever increasing quantity of genetically engineered protein therapeutics signals bioengineers to take stock of their alternatives for economic largescale isolation and purification of this new breed of pharmaceutical. Packed particulate chromatographic type columns, while suitable in laboratory and relatively smallscale production, often present scale-up problems. These problems derive frommediadeformation at high flow rates causing excessive pressure drops. The increasing need for reasonably high flow rates with tolerable pressure drops, while not compromising separation sharpness, is well known and has spurred commercial development along this path. The highly crosslinked, derivatized agaroses of Pharmacia (e.g., Fast Q ) and the prefusion technology of PerSeptive’s product line (8) are examples of packing media designed to permit higher flow rates. Silica-PVC (poly(viny1chloride)) microporous sheets, pioneered by FMC BioSupport Materials/BioProducts, attack the problem of media deformation by activating derivatized silica media that is embedded in a rigid PVC matrix. Additionally, the microporous sheet composite media can be stacked to achieve a desired height and, owing to structural integrity, allows for very squat geometries. The combination of squat geometry and tolerance to high pressure gradients results in a chromatographic material answering the need of high production rate with reasonable pressure drop. Another major advantage of the sheet format is that a variety of configurations are possible. For example, porous sheets, formed with fine parallel ribbing on the back, may be wound into a spiral channel (7). This channel provides
* Correspondence should be addressed to this author. t FMC Corporation.
* FMC BioSupport Materials.
for high flows of suspension-containing fluids. Proteins may be absorbed directly from biological suspensions into the porous sheet walls of the channel. BioSupport’s ActiMod Spiral Module is a commercially available unit. The cellulose porous sheets (CPS) to be discussed in this article are an out-growth of FMC BioSupport’s microporous sheet technology (1,2,9). This technology has been applied to chromatographic separation, bindrelease, and immobilized enzyme systems (3-6). The FMC BioSupport technology produces a porous sheet in which finely divided silica is embedded into a PVC support matrix. The silica provides a high surface area substrate which is suitable for derivatization with anionic, cationic, or other types of functional groups. This composite material has a wide range of uses. However, being silica based, it is not compatible with the caustic clean in place cycles which are often favored in commercial production. This article discusses porous sheets, which utilize the FMC BioSupport fabrication technology, but incorporate prederivatized cellulose rather than underivatized silica into a PVC matrix. The resulting porous sheet unites the following advantages: (1) accepts alkali clean in place cycles; (2) has low nonspecific binding; and (3) embeds standard, widely used, and well-characterized media such as Whatman’s DE ((diethylamino)ethyl), QA (quaternary amine), CM (carboxymethyl), SE (sulfoxyethyl), or P11 (phosphocellulose) media into a durable sheet format.
Materials and Methods Materials. The protein challengesolution incorporated bovine serum albumin (BSA,A-7906)or lysozyme (A-6876) from Sigma Co. dissolved in buffer. The two buffer solutions used were 0.01 M phosphate at pH 7.2 and 0.02 M acetate a t pH 4.7. Protein concentrations were 1mg/ mL.
8756-7938/93/3009-0285$04.00/0 0 1993 American Chemical Society and American Institute of Chemical Englneers
286
Bbtechd. Prog., 1993, Vol. 9,
No. 3
9
il
B F
r
.
H
20 pm
X-SECTION
40X
H 100 pm
600X
SURFACE VIEW
X-SECTION
60X
H
100 pm
Figure 1. Photomicrographs of DE32 CPS.
Phosphate buffer was prepared from anhydrousdibasic sodium phosphate, adjusted to the proper pH with phosphoric acid. Acetate buffer was prepared from anhydrous sodium acetate, and the pH was adjusted with acetic acid. Protein release solution was 1M NaCl in the buffer at the pH used for the challenge. The derivatized celluloses incorporated into the porous sheets were supplied by Whatman Paper Ltd., Springfield Mill, Maidstone, Kent, England. The Whatman codes and descriptions are as follows: Whatman code functionality DE32 (diethy1amino)ethyl QA52 trimethylhydroxypropyl CM32 carboxymethyl SE52 sulfoxyethyl P11 orthophosphate
type of exchanger weak base, anion strong base, anion weak acid, cation strong acid, cation strong and weak acid, cation
Binding Capacity. A description of the method for determining protein binding capacity follows. A 47 mm diameter disk was cut from a porous sheet. After the thickness of this disk was measured and its weight determined, it was placed in a Swinnex filter holder. The outlet of the filter holder was attached to an Isco (Lincoln, NE) on-line UV detector (ModelUA-5 with type 10optical unit), which measured the absorbance of the effluent a t 280 nm. Initially, buffer solution was pumped through the system to equilibrate the porous sheet and remove air bubbles. The flow was then switched to the protein challenge solution, which was pumped a t 5 mL/min (0.36cm/min). The effluent was collected during the pumping of this challenge solution. Pumping was continued until the optical density of the effluent reached 85% of the absorbanceof bovine serum albumin or lysozyme challenge
Blotechnol. Rug., 1993, Vol. 9, No. 3
287
Table I. Summary of BSA Binding Capacity Data for Cellulose Porous Sheets BSA capacity protein permeability binding (mgicm3 of released constant k media buffer pH bed volume) (%) (XlO'O cm2) 7.2 66 94 6.0 DE32" 0.01 M Phos 93 12.4 7.2 107 QA52" 0.01 MPhos 2.7 58 90 CM32" 0.02 M acetate 4.7 59 85 3.3 SE52Q 0.02 M acetate 4.7 100 15.4 0.02 M acetate 4.7 28b P11"
Pore Volume
2.50 2.00 1.50 1 .oo
0.50
Designation of Whatman Inc. In this case, the binding protein was lysozyme (at 1 mg/mL), not BSA.
0.00 0,001
0.01
0.1
100
10
1
solution (1mg/mL). At this point, the flow of protein challenge solution was stopped, and the volume pumped was recorded. Blank phosphate buffer solution was then pumped (at 5 mL/min) through the system until the optical density of the effluent returned to ita base-line value. The effluent during the challenge phase and the subsequent flush with buffer were collected in a single container, and its total volume was measured and recorded. After the collected effluent was stirred thoroughly, ita optical density was measured, and the content of protein was calculated. The protein bound by the disk was calculated by subtracting the amount of protein in the effluent from the amount of protein delivered to the disk. The second phase of the experiment determined how much of the bound protein was released from the porous sheet. Protein release solution (1M NaCl in buffer) was pumped through the disk at 3 mL/min. Effluent from this operation was collected in a clean vessel. Pumping was continued until after the optical density of the effluent had reached a maximum value and then returned to its base-line level. The total volume of effluent was measured, and after being mixed thoroughly, its optical density was noted. From these measurements the amount of protein in the effluent was calculated. Flow Characteristics. For a given phosphate buffer flow rate, the differential pressure across the column or the porous sheet (placed in a Swinnex holder) was measured byaCelescoDPcel1 (DP31-0300-111s/n22179). Porous Sheet Fabrication. Briefly, cellulose-based porous sheets were formed in a manner similar to that described in ref 1 and more specifically in ref 9 by (1) dry-blending PVC powder with the derivatized cellulose, (2) adding an organic solvent for the PVC such as cyclohexanone,(3) adding a swellingagent for the cellulose, Le., water, (4) blending this mix and extruding into asheet form, and (5) extracting the organic solvent from the sheet in a hot water bath and then drying the sheet.
Figure 3. Comparison of flow characteristics of apackedcolumn to a cellulose porous sheet: DE cellulose CPS; A,DE52 packed column.
Results and Discussion Binding Capacity. A summary of the BSA binding capacity of porous sheets which incorporate cellulose functionalities of DEAE, QA, CM, SE,and P11is contained in Table I. The sheets discussed in this article typically contain 58% by weight ion exchange media unless otherwise noted. However, media content can be varied over a range of at least 33 5% to 65 % . The protein challenge solution was either BSA or lysozyme at a concentration of 1mg/mL in a phosphate or acetate buffer. Most of our work focused on the DE cellulose. Porous sheets containing this functionality have a BSA binding capacity of about 66 mg/mL. This is about 66% of the capacity of a column packed with neat particulate DE52. Recovery of protein from the porous sheet typically measured in the 90-100% range (see Table I). Also given in Table I are the permeability constants associated with
each of the embedded media/PVC composites. Permeability will be discussed more fully in a later section. Sheet Morphology. Photomicrographs depicting the top and cross-sectional views of a CPS containing 61.5% Whatman DE32 are shown in Figure 1. It is seen that noodles of the cellulose media are housed within the swiss cheese structure of the PVC matrix. Photomicrographs were obtained with an ETEC electron microscope interfaced with PGT's IMEX computerized image and microanalysis process systems. Pore Size Distribution. Mercury intrustion measurements of the CPS composites were made with a Quantachrome (Syosset, NY) Autoscan-33 porosimeter. Intrusion pressures ranged from 0.5 to 33 000 psia. Typical data for a CPS containing 58.3% Whatman DE32 are shown in Figure 2. Pores span the range 20-40 pm, with an average of 23 p m . Note that these measurements are performed on dry material and therefore do not
Pore Diameter ( p) Pore Size Dlstribution
0.35 0.30
1
I
I
1
10
0.25
0.20 0.15 0.10
0.05 0.00 100
Pore Diameter ( pm)
Figure 2. Pore size data for DE32 cellulose porous sheet. 120.0
I
I
I
I
I
I
I
1
0 01 M Phosphate Buffer 0
50
100
.,
150
200
250
300
350
400
Pressure Gradient (PSllcm)
Biotechnol. Prog., 1993, VOl. 9. No. 3
288 1oc
eo
60
I N
0
40
20
0 ! art Phos.
Buffer
Start Egg White Challange Solution
Desc: Single Disk 47 mm Buffer A: 0.01 m Phos., 7.2 pH Buffer B: A + 0.5 M NaCl Test Mixture: Fresh Egg White Conc. (mg/ml): 14.3 Amount (ml): 1.1 Gradient: Linear, 0-100% B Total Voi. (mi): 200
t
Start NaCl Gradient Elution
MINUTES
Flow Rates (mi/min) (cm/hr) 5 21.7 Binding: Flush: 5 21.7 Elution: 2 8.7 T U Time (min): 89 Recorder ISCO Speed (cmihr): 6 Sensitivity: 0.1
diameter is ca
n
Test autnrc: Natural
C
e 0.20-
-0.2
H 1
egg vhite, 100 mg protein iu 7.0 d of blrffer. Pd(a)1p327me, -
289
Biotechnol. Frog., 1993, Vol. 9, No. 3
Table 11. Theoretical Comparison of DE32 CMPS to DE52 Packed Columna DE52 DE32 DE52 DE32 packed CMPS packed CMPS column module column module 220 2330 14.6 Q, mllmin diam, cm 4 56 5OC 1 APIL, psilcm ht, cm 10 13.9 167 velocity, cm/min 17.5 vol, cmR 126 7Eid 6 50 k X lolo,cm2 AP,psi 50
Table 111. Caustic Cleaning Cycle
flush of pH 7.5 binding flush elution NaOH cleaning wash
solution 0.01 M Phos 0.01 M Phos + BSA 0.01 M Phos 1M NaCl + 0.1 M Phos 0.5 N NaOH
mL/ min 5.0
5.0 5.0
approx time (min) 20 55
2.0
35 40
5.0
4
The requirements were they had to have the same protein binding capacity and the same pressure drop. * Can go to maximum of 11 psi/cm. Can go to maximum of >375 psilcm. d Depends on packing; can be considerably lower. (I
material. The column was packed with DE52 and had dimensions of 1.6 (diam) X 1.6 cm (height). The porous sheets (several were studied), containing Whatman DE32, were about 0.045 in. thick and 13-47 mm in diameter. The flow characteristics for the packed column and the porous sheet are depicted in Figure 3. Notice that in the velocity range 0-20 cm/min and pressure domain 0-11 psi/cm the packed column has a significantly higher permeability than the CPS. That is, the packed column permits higher linear superficial velocities at the same pressure gradient. However, the packed column reaches an upper limit of velocity of 20 cm/min corresponding to a pressure gradient of 11psi/cm. At this pressure gradient, the packed column squashes together (the packing is physically compressed, being about 90% of its original height) and the velocity levels off (see Figure 3). Thus, the packed column will not permit a superficial velocity in excess of 20 cm/min. In contrast to the upper velocity of the column of 20 cm/min, the CPS material will allow in excess of 100 cm/min. At 100 cm/min the pressure gradient supported by the CPS was about 400 psi/cm. A theoretical comparison is now made of a packed DE52 column to a module of stacked porous sheets containing Whatman DE 32 (see Table 11). The constraints are that the protein binding capacity and applied pressure drop for the column and the module be the same. The packed column is to have dimensions of 4 (diam) X 10 cm (ht). Typically such relatively slender aspect ratios are used to insure good flow distribution through the packing. By contrast, the stacked sheet module is short andsquat, being 14.6 (diam) X 1cm (ht). The porous sheet, owing to its lower permeability and its structural strength, can be fashioned to such a low aspect ratio. From Table 11, note that the equivalent short squat module of stacked sheets can achieve more than 1OX the flow rate of the column (2330 mL/min vs 220 mL/min). Hence, it is possible to design the stacked sheets to get more than 1OX the productivity from a module compared to an equivalent packed column. This increase in flow rate was achieved primarily through the squat geometry of the stacked sheets. Furthermore, the pressure applied to the CPS module could theoretically be increased by a factor of 8.0 (to obtain a pressure gradient of 400 psi/cm). The applied pressure gradient to the packed column could be approximately doubled. This would, in theory, give the module a 40-fold advantage in productivity over the packed column. Chromatography by a Single Sheet. Salt gradient elution chromatography was performed on natural egg white protein using a single porous sheet containing 64 76 DE32. The single sheet was 1.3 mm thick and 47 mm in diameter. The chromatograph and pertinent data are shown in Figure 4A. A nice separation of lysozyme, conalbumin, and ovalbumin is evidenced. For comparison purposes, a chromatograph from Whatman's literature of natural egg white obtained with a
Figure 5. Traces from data acquisition systems during BSA binding to DE CPS.
conventional packed column (1.5 (diam) X 10cm (length)) of DE52 is shown in Figure 4B. Evidently, a level of chromatograph separation comparable to a 10 cm long column is possible on a sheet of only 0.13 cm thickness. Effect of Caustic CleaningonCellulosePS. A single 47-mm disk (ca. 1.55 mm thick) of DE32 was subjected to 10 cleaning cycles. The several components of each cycle are shown in Table 111. Thus, the disk was subjected to 4 min of 0.5 N NaOH per cycle. The approximate times, except for the NaOH wash which is exact, for each portion of the cycle are shown in Table 111. The times were approximate as, for example, the binding portion was terminated when 85 96 of total absorbance was reached or, during the buffer flush, when base line was achieved. No loss of BSA binding capacity after 10 cycles was observed, being undiminished at about 130 mg on a 47 mm diameter disk. The 0.5 N NaOH was found to reduce the pressure drop through the CPS. Evidently, the caustic solution can cause shrinkage of the cellulose fibers. These data indicate that a cellulose porous sheet is capable of withstanding caustic clean in place cycles. Comments on Binding Cycle Behavior. A Keithly 575 data acquisition system was used to continuously monitor effluent OD at 280 nm, pressure drop, and conductivity during a typical BSA binding run on a DE32 sheet. A graph of these parameters is shown in Figure 5. This plot gives a sense of the behavior of the individual parameters. It may be noted that the pressure drop through the disk gradually builds during the binding cycle, remaining constant at its maximum during the buffer flush. For the case in Figure 5, the pressure drop increased from ca. 0.15 psi before binding BSA to 0.55 psi near full binding of BSA. Sheet thickness was about 1.2 mm. Upon switching from the phosphate buffer to the 1 M NaCl elution solution, the pressure drop across the sheet falls to its base-line value. These data demonstrate that a bound protein can cause an increased pressure drop through the binding media.
Blotechnol. Prog., 1993, Vol. 9, No. 3
290
Acknowledgment The authors are indebted to R. Y. Chen, S. Nochumson, and S. Sloshberg for their many helpful suggestions on this project. Thanks also to A. J. Young, N. Lieu, and W. Dodge for their excellent technical support in obtaining the data reported here. Literature Cited (1) Selsor, J. Q.;Turner, E. W.; Goldberg, B. S. U.S. Patent No. 3,696,061,October 3, 1972. (2) Goldberg, B.S.U.S.Patent No. 3,862,030,January 2,1975. (3) Goldberg, B. S.U.S. Patent No. 4,102,746,July 25, 1978. (4) Goldberg, B. S.U S . Patent No. 4,169,014,September 25, 1979. (5) Goldberg, B. S.; Hausser, A. G.; Gilman, K.; Chen, R. Y. Immobilized Microbial Cells. In Evaluation of a Novel
Microporous PVC-Silica Support for Immobilized Enzyme; Venkatsubramanian, K., Ed.; ACS Symposium Series 106; American Chemical Society: Washington, D.C., 1979;pp 173186. (6) Hausser, A.G.; Goldberg, B. S.; Mertens, R. An Immobilized Two-Enzyme System (Fungal a-Amylase/Glucoamylase)and Its Use in the Continuous Production of High Conversion Maltose-Containing Syrups. BiotechnoL Bioeng. 1983,25,525539. (7) Goldberg, B.S.;Chen, R. Y. U.S. PatentNo. 4,689,302,August 25, 1987. (8) Afeyan, N. B.;Fulton, S. P.; Regnier, F. E. High Throughput Chromatography Using Perfusive Supports. LC-GC 1991,9, 824-832. (9) Manganaro, J. L.;Goldberg, B. S.; Raynor, G. E.; Gray, C. A. U S . Patent No. 5,155,144,October 13,1992. Accepted September 21, 1992.
