Journal of Biotechnology, 26 (1992) 213-229
213
© 1992 Elsevier Science Publishers B.V. All rights reserved 0168-1656/92/$05.00
BIOTEC 00789
Process development for the recovery and purification of recombinant protein G S h w u - M a a n L e e , N i n a Z. Essig and T i m o t h y K. L e e Genex Corporation, 16020Industrial Drive, Gaithersburg, MD 20877, USA
(Received 22 April 1991;revisionaccepted 2 March 1992)
Summary The domains of protein G from streptococcus which bind immunoglobulin G have been cloned and expressed in Escherichia coli (Fahnestock et al., 1986). Because protein G binds to several animal immunoglobulin G's, it has many immunochemical applications. This report describes process development for large-scale production of this recombinant protein G (also known as GammaBind ® G). In 200 1 cultures of E. coli, this protein G variant was released from the cell into the culture medium by heating at 80°C for 10 min. The concentration was monitored by either a competitive enzyme-linked immunoassay or a liquid chromatographic assay. Cross-flow microfiltration with 0.22 Ixm membrane was used to remove the cells. The protein G-rich permeate from the crossfiow microfilter was purified by affinity chromatography using a 5 1 column of IgG-Sepharose 6 Fast Flow, which yielded 16-18 g of protein G per column cycle. The pools of purified protein G were concentrated and desalted using ultrafiltration. The salt-free protein G was then lyophilized as bulk product. The overall recovery through the entire process was 50-64%. The analysis of the final product included sodium dodecyl sulfate polyacrylamide gel electrophoresis, UV-visible spectrum, high performance gel filtration, endotoxin level and binding efficiency to human IgG Sepharose. Protein G; Cross-flow microfiltration; Downstream processing; Affinity chromatography; Ultrafiltration
Correspondence to: S.-M. Lee, AMVAX, Inc., 12040 Indian Creek Court, Beltsville, MD 20705, USA.
214 Introduction
Protein G is a cell wall protein of group G streptococci (Bj6rck and Kronvall, 1984; Boyle and Reis, 1987). Like protein A of Staphylococcus aureus, protein G binds to the Fc domains of a wide variety of immunoglobulin G's (IgG's). The potential applications of protein G include virtually all of the current and projected applications of protein A. These can be grouped into three categories: (a) the detection and quantitation of antibodies and immune complexes; (b) the purification of antibodies and immune complexes; (c) the removal of antibodies and immune complexes from serum of patients by extracorporeal plasma filtration. There has been considerable interest in using protein G in applications for which protein A is not well suited due to specificity differences. In general, protein G is more specific than protein A in that it binds only to IgG and not, for example, to IgM which protein A binds. On the other hand, protein G shows broader IgG reactivity, binding several important IgG's which protein A does not bind well. These include human IgG3, mouse IgGa, goat and rat IgG's. Until now, protein G has not been widely available because it is difficult to prepare from streptococci. The gene has been cloned from a streptococcus clinical isolate into Escherichia coli and Bacillus subtilis (Fahnestock et al., 1986; Fahnestock, 1987). The recombinant protein G (also known as GammaBind ® G) in this report refers to a modified form of protein G constructed by recombinant techniques and expressed in E. coli; for simplicity, the name protein G is used throughout this report. This recombinant protein G contains the two IgG binding B domains followed by a sequence rich in proline residues and a set of five 'C-repeats', analogs of the pentapeptide Asp-Asp-Ala-Lys-Lys, near the C-terminus (Fahnestock, 1987). The deleted sequences include those which are responsible for albumin binding in protein G. This report describes the assay development and process development for large scale production of this recombinant protein G.
Materials and Methods
Competitive enzyme-linked immunosorbent assay (ELISA) for protein G determination Protein G-horseradish peroxidase (HRP) conjugate was prepared according to the periodate-oxidation method (O'Sullivan and Marks, 1981; Nakane and Kawaoi, 1974). The weight ratio of HRP to protein G used in conjugation was 20 to 1; the conjugate mixture was not separated from unconjugated proteins. Microtiter plates (Immulon 2, Dynatech) were coated at room temperature overnight with 110 ~xl (110 ng) of human IgG (Jackson Immunoresearch Lab) in 50 mM sodium phosphate buffer containing 0.15 M NaC1, pH 7.4. The coating solution was removed and the plates were blocked by incubation for 1 h or longer at room temperature with 200 Ixl of casein (Fisher), which had been prepared as a 2% solution in 7 mM sodium phosphate buffer, pH 7.4, containing 0.1 M NaCI and 0.2% sodium azide. Before sample and standard were added, the plates were
215 washed once with 10 mM sodium phosphate buffer, 0.15 M NaC1, pH 7.4, containing 0.05% Tween 20 (Sigma). Protein G standard was prepared by affinity purification through IgG Sepharose 6 Fast Flow and its concentration determined using an extinction coefficient of 1.0 for a 1 mg m1-1 solution at 280 nm. The standard and samples were prediluted in the tubes, followed by a twofold serial dilution in the plates and incubated for 1 h. To the plates containing 100 ixl of sample or standard, 10 txl of protein G - H R P was added and the incubation proceeded for 30 min. The conjugate was prepared by diluting a stock solution of the conjugate (protein G concentration at 0.06 mg m l - l ) 5000-fold with 50 mM sodium phosphate buffer, 0.15 M NaC1, pH 7.4, containing 0.1% bovine serum albumin (Sigma, RIA grade) and 0.05% Tween 20. The same dilution buffer was used for both standard and samples. The plates were washed ten times after the conjugate incubation and the color was developed with the 3,3',5,5'-tetramethyl benzidine substrate system (Kirkegaard and Perry Lab). The color reaction was stopped at 30 min with 1 M phosphoric acid. The absorbance at 450 nm was measured with a multiscan spectrophotometer (Titertek).
Liquid chromatographic assay for protein G determination Liquid chromatography (LC) was performed on a Waters liquid chromatography system consisting of a model 720 controller, a model 730 data module, a model 6000A pump, a M-45 pump, a model 440 absorbance detector, a U6K injector with a 2 ml loop and an analytical column of immobilized human IgG. Crude protein G sample or affinity purified protein G standard at pH 6.5-7.5 was loaded at a flow rate of 1 ml min-1 onto the column which had been equilibrated with 45 mM potassium phosphate buffer containing 150 mM NaC1 and 0.005% (w/v) sodium azide, pH 7.4 (PBS/NaN3). The column was then washed with the same buffer to remove unadsorbed material at a flow rate of 5 ml min-1. The adsorbed protein G was eluted with 0.5 M ammonium acetate, pH 3, at a flow rate of 5 ml min-1. Following elution, the column was re-equilibrated with P B S / N a N 3. The run time was 20 min. From the chromatogram, the peak height was matched to a standard curve to determine the amount of protein G present in the sample.
Recovery of protein G from spent medium At the end of an E. coli cultivation, protein G was released from the cells into the spent medium by heating the bioreactor at 80°C for 10 min (Finkelman and Lee, 1990). During the early phase of process development, the spent medium was often aged at 5°C for 1-3 days after the heating process to facilitate the release of protein G from the cells. The cells were separated from the protein G-containing spent medium by cross-flow microfiltration (CFM) through a Prostak-1 pilot unit (Millipore) using a 4.6 m 2 membrane (GVPP type, 0.2 ~m, wide channel, Millipore) operated at ambient temperature. Initially the system was drained and the integrity of the membrane was tested in the system by measuring air diffusion rate at 2 bar to be less than 14.5 ml m -2 min -1. Once the integrity of the membrane was confirmed, the cells in the spent medium were concentrated 10-20-fold across the membrane
216
and the protein G-rich permeate was collected. The concentrated cells (retentate) were then washed continuously with 100-200 1 of P B S / N a N 3 on the Prostak membrane. At the end of CFM, the retentate was discarded and the system was cleaned in accordance with the manufacturer's instruction which included saline wash, water wash, detergent wash, sodium hypochlorite sanitization and warm acetic acid wash. Following a water wash, the system was stored in 0.05% sodium azide.
Purification of protein G IgG Sepharose 6 Fast Flow (Pharmacia) was suspended in 20 mM Tris buffer, pH 7.5, and packed in a 5-1 column (25.2 cm × 10 cm, Pharmacia) at 5°C. The flow was maintained with a peristaltic pump (Watson-Marlow type 604 U / R , Bacon Technical Industries, Inc.). Two on-line filters (1 and 5 txm Polycap TM 75HD, Whatman) were used to protect the column from particulates and a pressure gauge was installed between the pump and the filters. The column was initially washed with 20 mM Tris buffer, pH 7.5, to remove MgClz and ethanol (preservatives) and subsequently washed with 1 M ammonium acetate, pH 3, to remove unbound lgG from the gel matrix. P B S / N a N 3 was prepared as a fivefold concentrate and clarified by passing through a 1/xm polypropylene cartridge filter (Pall). The column was equilibrated with PBS/NaN3, the output from Prostak was loaded onto the column in an upward direction using a linear flow rate of 150 cm h-1 (75 I h-1), which produced an operating pressure of 1 bar. The column was washed with P B S / N a N 3 (75-150 1) until the absorbance at 280 nm dropped to baseline as indicated by the on-line UV monitor (Type UA-5, ISCO). The column was subsequently washed with 5 mM ammonium acetate buffer, pH 5.0, (25-50 1) to remove the bulk of the salt from the column and bring the pH down to 5. The column was eluted with 0.5 M ammonium acetate, pH 3, in a downward direction using a linear flow rate of 75 cm h-1. This protein G solution was adjusted to pH 5 with concentrated ammonium hydroxide and stored at 5°C. The column was equilibrated again with P B S / N a N 3 and stored in the same buffer at 5°C.
Ultrafiltration The purified protein G was pooled for concentration and dialysis. A Pellicon system (Millipore) with a 0.46 m 2 polysulfone membrane cassette (PTGC type, 10000 nominal molecular weight cut off, Millipore) was used. The flow was controlled by a peristaltic pump (Watson-Marlow type 604 U / R ) . The membrane cassette was integrity tested at 0.3 bar inlet pressure and confirmed that the air diffusion rate was less than 10 ml rain-1. The ultrafiltration system was operated at a recirculation rate of 4 1 min-1 at room temperature. Following the initial concentration, further dialysis was conducted by batch addition of equal volumes of 2 mM N H a H C O 3 buffer, pH 7. The desalting process was monitored by measuring the conductivity of the permeate. The system was cleaned by recirculation of 0.1 N NaOH. Following a water wash to remove the residual NaOH, the system was stored in 0.05% NaN 3.
217
Lyophilization The salt-free protein G solution was clarified by 0.22 I~m filtration (Millipak 60, Millipore) and lyophilized in bulk using a Freeze Mobile 24 (Virtis).
Sodium dodecyl sulfate polyacrylamide gel electrophoresis (SDS-PAGE) SDS-PAGE was performed according to the method of Laemmli (1970). Gels of 1.5 mm thickness were constituted with 15% polyacrylamide and stained with Coomassie blue. Molecular weight protein markers (Bio-Rad) included phosphorylase b, 97 kDa; bovine serum albumin, 67 kDa; ovalbumin, 43 KDa; carbonic anhydrase, 31 kDa; soybean trypsin inhibitor, 22 kDa; and lysozyme, 14 kDa.
Protein determination The concentration of affinity purified protein G was determined by its UV absorbance using an extinction coefficient of 1.0 at 280 nm for a 1 mg m1-1 (0.1%) solution. The protein concentration in the crude protein G preparation was determined using the Pierce bicinchoninic acid assay (Smith et al., 1985). Affinity purified protein G was used as a standard to construct the calibration curve.
HPLC gel filtration assay for purity determination We used a TSK G2000SWXL column (0.78 cm x 30 cm, TosoHaas) with its guard column (0.6 cm x 4 cm). The running and column storage buffer was 0.1 M sodium phosphate containing 0.1 M sodium sulfate and 0.05% sodium azide, pH 6.4. The flow rate used was 0.5 ml min-1 and the total run time was 30 min.
Binding efficiency assay An IgG Sepharose 6 Fast Flow column (1.6 cm X 3 cm) was pre-eluted with 1 M acetic acid and equilibrated with P B S / N a N 3. The flow rate was maintained at 0.6 ml min-1 with a peristaltic pump. Lyophilized, salt free protein G was dissolved in P B S / N a N 3 to a final concentration of 1.2 mg m1-1 and a total of 5 ml was applied to the column. Following the column wash with P B S / N a N s , the protein G bound to the column was eluted with 1 M acetic acid. The 280 nm absorbance of each individual fraction was measured using a spectrophotometer (Shimadzu, UV-260). The binding efficiency was calculated according to the following formula: A280 eluted Binding efficiency (%) = A2s0 in breakthrough and wash + A2s 0 eluted
Endotoxin assay Endotoxin was determined using the gel-clot method performed by Associates of Cape Cod, Inc. Woods Hole, MA. A total of 9 mg of lyophilized protein G was reconstituted with 1 ml of endotoxin free water and titered down using a twofold dilution scheme. E. coli endotoxin standard was added to sample dilutions as inhibition controls. The assay sensitivity was 0.03 E U ml-1.
218
Results Protein G assay
In the initial process development stages, the competitive ELISA was used to monitor protein G concentration in the crude sample. In this assay, protein G competed with protein G - H R P conjugate for binding to human IgG adsorbed onto microtiter plates. The working range was 30-1000 ng m1-1 (Fig. 1). The assay was adequate to determine protein G concentration in the spent medium following heat release. However, the major drawback was its speed as it usually required 4 h to complete one assay. This assay was therefore impractical to use for in process monitoring. The specific binding between protein G and human IgG as well as the availability of an immobilized IgG analytical column allowed us to establish an LC assay. Figure 2 shows a typical LC chromatogram. The assay was validated by adding a protein G standard to a typical crude protein G sample (i.e., the Prostak output) from which protein G had been specifically removed by IgG Sepharose 6 Fast Flow. The standard curve remained the same in the presence of the contaminants in the Prostak output (Fig. 3). It was concluded that there was minimal interference from the contaminants and that the eluted peak in the LC assay indeed represented the amount of protein G in the crude sample. The capacity of the immobilized IgG analytical column was determined to be 200 Ixg. A convenient working range was 20-160 I~g of protein G per assay. With the 2 ml injection loop the detection limit was 10 Fxg m l - 1. This assay was adequate to support the process development work. The immobilized IgG analytical column had an average life of six months or 300-500 injections under our operating conditions. As the column deteriorated,
2.
[ o I/1
W C in I. O
0
. . . . . . . .
1o
f
1oo
. . . . . . . .
i
1ooo
. . . . . . . .
1oooo
Protein G (ng ml -1) Fig. 1. R e s p o n s e of c o m p e t i t i v e E L I S A to various c o n c e n t r a t i o n s of p r o t e i n G s t a n d a r d . T h e points on the curve r e p r e s e n t d u p l i c a t e v a l u e s p e r f o r m e d on the s a m e plate.
219
IA(280nm) Loacl
Wash
[
Elutlon
I
Reequlllbratlon
I
110
1
I
15
Time (mln) Fig. 2. Immobilized IgG analytical column loading and elution profile. A total of 300 ~1 of standard protein G at a concentration of 0.48 mg m l - 1 was manually injected onto the column. The flow rate was 5 ml m i n - 1 except during the loading stage when it was 1 ml min 1. The full scale absorbance was 0.2 at 280 nm.
the peak height dropped and the assay sensitivity was diminished. The advantages of this assay included short analysis time (20 min), relative simplicity and reproducibility compared to the ELISA.
15
10 v u
~t
5
Q,.
0 0
100
200
Protein G (/Jg) Fig. 3. Standard curve obtained by injecting known amounts of protein G onto an immobilized IgG analytical column. Affinity purified protein G solution at a concentration of 0.48 mg m l - I was applied to the column at volumes of 100-400 ixl (e). The same standard was diluted 5-fold with Prostak output from which the protein G had been specifically removed by IgG Sepharose 6 Fast Flow and 0.5-2 ml of the diluted standard was loaded onto the column (o). Elution peak height was plotted as a function of the a m o u n t of protein G loaded.
220
Solid-liquid separation The solids content in the protein G-released spent medium was usually about 2.5% as measured in a Gyrotester (Alfa-laval). This relatively low solids content made CFM an attractive approach for initial solid-liquid separation. Initially, the Prostak unit was operated at a cross flow rate of 170-190 1 min-1 which produced an inlet pressure of 3.4-3.9 bar and a retentate pressure of 0.8-1.0 bar. The p e r m e a t e and retentate valves were fully opened. The average transmembrane pressure (TMP) was calculated to be 2.1-2.5 bar according to this formula: T M P = (inlet pressure + retentate pressure) - 2. Using these operating parameters, the p e r m e a t e flow rate was measured to be 2.8-3.5 1 m i n - 1 which was equivalent to 37-46 1 m -2 h -x (LMH). The percentage of protein G transmitted through the m e m b r a n e was 39% initially and gradually declined to 20% when the cells were concentrated sixfold. The protein G recovered in the initial 10-fold concentration of the cells was less than 50%. In combination with 180 1 of cell washing, the yield was about 75%. The operating parameters had to be modified to improve the transmission and yield. The cross flow rate was reduced to 135 1 min -~ which produced an inlet pressure of 2.6-2.9 bar. The retentate gauge indicated 0.7-0.9 bar of pressure with the valve fully opened. A diaphragm valve was installed in the p e r m e a t e stream to adjust TMP, which was calculated according to the following formula: T M P = (inlet pressure + retentate pressure) + 2 - permeate pressure A p e r m e a t e pressure of 0.7 bar was used which produced a T M P of 1.0-1.2 bar. Figure 4 shows the transmission and p e r m e a t e flux along the primary concentration process using the modified procedure with 1.0-1.2 bar TMP. When compared to the original operating parameters with higher TMP, about twice as much protein G transmitted through the m e m b r a n e with only a slight reduction of p e r m e a t e flux. Table 1 shows the result of a typical run; 72% of the input protein G was recovered in the p e r m e a t e through the initial cell concentration step and 28% through the washing step. The final retentate contained only 2.7% of the input protein G and the yield was quantitative when sufficient wash was used. Without the washing step, we expected a 72% yield.
Affinity chromatography In the m o d e of packed column operation, the sample is fed continuously into the column until the available capacity of the column (IgG Sepharose 6 Fast Flow) is exhausted and the adsorbate (protein G) begins to appear in the column effluent. The variation of the concentration of adsorbate in the column effluent as a function of the amount of adsorbate applied to the column is a breakthrough curve (Chase, 1984). To understand the interaction between protein G and I g G Sepharose and to optimize the affinity process, breakthrough curves were established in a laboratory column (1.6 cm x 10 cm) which was kept at the same length as the pilot column (25.2 cm × 10 cm). The cross-sectional areas for the pilot and laboratory columns
221
50
100
A
90"
~
Transmission • 40
A "r"
80'
--1 v • 30
E ¢n
70@ • 20
60
E II.
10 a.
50
0
40
10
Concentration
factor
Fig. 4. Protein G transmission and permeate flux during the course of cell separation in the Prostak under 1.0-1.2 bar TMP. Concentration factor is the ratio of original volume to retentate volume at various points along the cell concentration process. Protein G transmission is the ratio of protein G concentration in the permeate to protein G concentration in the retentate. The LC assay was used for protein G determination.
were 500 and 2 cm 2, respectively. Therefore, the scale-up factor in terms of column bed volume and column capacity was the ratio of the two cross-sectional areas, i.e., 250. The Prostak permeate with protein G concentration adjusted to 0.18 mg m1-1 was continuously fed into the IgG Sepharose column (1.6 cm x 10 cm) until the available capacity of the column was totally exhausted. A linear flow rate of 150 cm h-1 was used. The breakthrough curve was constructed by plotting the percentage of protein G appearing in the breakthrough fractions against the amount of protein G loaded (Fig. 5). The yield curve was established by plotting the
TABLE 1 Protein G recovery through Prostak pilot-1 unit
Protein G-containing spent m e d i u m Permeate collected through initial cell concentration Permeate collected during cell washing stage Final retentate T M P of 1.0-1.2 bar was used.
Normalized protein G concentration
Volume (1)
Yield (%)
1.0 0.76
190 180
100 72
0.28
185
28
0.25
20
2.7
222
100
o~"
• 100
80
' 80
40
.40
o
.S
~.
o 20
rough
/ 0
n.~ 50
,20
Yield 0 100
1 50
200
Amount of protein G loaded (rag)
Fig. 5. Breakthrough curve and yield curve at a linear flow rate of 150 cm h -1. Prostak output with protein G concentration adjusted to 0.18 mg m1-1 was loaded onto an IgG Sepharose 6 Fast Flow column (1.6 cm × 10 cm) at 5 ml m i n - 1. The LC assay was used to determine protein G concentration in the breakthrough fractions and the column starting material. Protein G eluted from the column was determined by UV absorbance at 280 nm.
percentage of protein G recovered against the different amounts of protein G loaded (Fig. 5). The column's saturation capacity was determined to be 4 mg of protein G per ml of gel. The laboratory column's full capacity was 80 mg. The breakthrough curve in Fig. 5 indicated that when 56 mg of protein G was applied to the coltrmn, protein G started to leak from the column and the percentage leaking increased as the loading amount increased. Beyond this breakthrough point, the more protein G loaded, the more protein G would be lost in the breakthrough and the lower the yield. In a production situation, it would be the best to utilize an affinity column's full capacity so as to decrease the number of production cycles. However, to completely saturate the column, more than 160 mg of protein G (twice the column capacity) had to be applied and the yield would be lower than 50%. Therefore, if the IgG Sepharose column was to be fully utilized with a reasonable yield, the protein G lost in the breakthrough had to be recycled. Applying 70-90 mg of crude protein G to the laboratory column, which was equivalent to 18-23 g to the pilot column, appeared to be a better compromise for routine operation; the yield under these conditions was predicted to be 65-75%. With this loading amount, there was a small loss in the breakthrough, but most of the column capacity was utilized. In an affinity chromatography system, the most efficient adsorption performance is obtained when the deflection point of the breakthrough curve is the sharpest. The concentration of adsorbate and loading flow rate can both affect the shape of the breakthrough curve (Chase, 1984). The effect of loading flow rate was
223
80
80 A
-1
v .C
6O
a
-
150 cm h
•
75 cm
,60
1
4O
.40
2O
.20
.E 0
i
O.
50
100
150
200
Amount of protein G loaded (rag) Fig. 6. Breakthrough curves at linear flow rates of 150 and 75 cm h - I. The starting material for the IgG Sepharose column (1.6 cm × 10 cm) was affinity purified protein G at concentration of 0.33 mg m l - 1 . T h e protein G concentration in the breakthrough was determined by U V absorbance at 280 nm.
assessed by loading purified protein G onto the laboratory column at 150 and 75 cm h - t (Fig. 6). Applying 70-90 mg of protein G to the laboratory column with a saturation capacity of 80 mg, one would expect a slightly higher yield with a slower flow rate as indicated in the breakthrough curves. However, the slight improvement in yield with the slower flow rate did not justify spending additional production time in loading. The effect of adsorbate concentration on the shape of the breakthrough curve was also investigated. For this experiment, the Prostak permeate was adjusted to a final protein G concentration of 1.1 mg ml-1. When this material was compared to that of 0.18 mg ml-1, the two breakthrough curves overlapped (Fig. 7). Although, concentrating the Prostak output reduced the loading time, there was no improvement in the adsorption efficiency and yield. Table 2 presents data from the pilot affinity column. As predicted from the development work, the first half of the breakthrough contained no protein G and the second half contained a small amount. In this example, 83% of the column capacity was utilized and the yield was 72%.
Ultrafiltration The ultrafiltration process was optimized by conducting a pressure excursion experiment. The permeate flow rate was measured at 0.14 bar increments of average TMP [(inlet pressure + retentate p r e s s u r e ) - 2]. The optimized average TMP was determined to be 1.3 bar, which was the point where increasing TMP did not increase the flux. The permeate flow rate achieved initially was 0.26 1 min-1 and gradually slowed down to 0.15 1 min -1. Protein G was concentrated 28-fold with 87% recovery, to give a solution of 46.2 g protein in 2.1 1.
224
A
80
- 80
60
-60
40
"40
20
"20
=,..
o
.,Q
0 Q.
•
20
40
60
80
100
!
120
140
Amount of protein G loaded (rag) Fig. 7. Breakthrough curves established at a linear flow rate of 150 cm h - i using two different starting materials with protein G concentrations of 0.18 and 1.1 mg ml - I . Protein G was determined by LC assay.
Overall recovery The yield for initial solid-liquid separation was 90-98% depending on the amount of wash used. The yield for affinity chromatography was 65-75% depending on the amount of protein G loaded. The yield through ultrafiltration step was generally 85-87%. Therefore the total yield throughout the entire process was 50-64%.
Analysis of affinity purified protein G Affinity purified protein G was characterized by its absorbance maximum at 278 nm, absorbance minimum at 250 nm and a distinct shoulder at 283 nm. The A280/A260 ratio was calculated to be 1.77-2.00 in several lots tested. According to Layne's method (1957), an A280/m26 o ratio of lower than 1.75 indicates significant nucleic acid contamination of a protein solution. Therefore, the affinity purified protein G was low in nucleic acids.
TABLE 2 Protein G recovery from pilot affinity column Total protein G
Yield (%)
(g) Starting material (Prostak output) Breakthrough fraction 1 Breakthrough fraction 2 Column wash Elution
23.0 0 1.9 1.3 16.6
100 0 8 6 72
The bed volume of the column was 5 1 (25.2 cm × 10 cm) and flow rate was set at 150 cm h - i (75 1 h - 1).
225
1
2
3
4 ---97 K "-'67 K ---43K ---31 K
---22K --14 K
Fig. 8. SDS-PAGE of crude and purified protein G. Lane 1: Prostak output containing 20 txg of protein G; Lanes 2 and 3: affinity purified protein G, 20 and 1 I~g, respectively; Lane 4: molecular weight standards. F i g u r e 8 s h o w s t h e S D S - P A G E o f c r u d e a n d p u r i f i e d p r o t e i n G. S i n c e p r o t e i n G w a s r e l e a s e d f r o m t h e cells w i t h o u t m e c h a n i c a l b r e a k a g e , t h e l e v e l o f p r o t e i n c o n t a m i n a n t s in t h e s p e n t m e d i u m was v e r y low. T h e p r e d o m i n a n t c o m p o n e n t in
~o
vI d
A
Time (min) Fig. 9. HPLC gel filtration analysis of affinity purified protein O. Lyophilized protein G was reconstituted in 0.1 M sodium phosphate, pH 6.4, containing 0.1 M sodium sulfate and 0.05% sodium azide to a final concentration of 1.4 mg ml- 1. A total of 15 p,l was injected and the elution was monitored at 280 nm. The full scale absorbance was 0.1. V0 indicates the void volume determined by blue dextran and Vi is the included volume determined by p-aminophenylalanine.
226
0.05
Load
Wash
0.5 I
Elution
A (280 nm)
I
2 T i m e (h)
Fig. 10. U V monitor tracing from binding efficiency assay. The full scale absorbance was set at 0.1 for loading and washing stages and set at 1.0 for elution. The baseline was readjusted upon switching to 1.0
absorbance. the Prostak output was protein G with only a few minor impurities (lane 1). Protein G purified from the current procedure appeared as essentially a single major band on the SDS gel (lanes 2 and 3). SDS-PAGE was a sensitive way to detect contaminating proteins, and had been extremely useful in monitoring our process development work. However, it was difficult to quantitate protein G purity reproducibly based on scanning of stained gel. The HPLC gel filtration assay offered a more quantitative method for determining protein G purity. As shown in Fig. 9, two minor impurities were detected using this method. The purity was calculated to be 97% using integration of the peak areas. The affinity purified protein G was reapplied to the IgG Sepharose Fast Flow column as described in the binding efficiency assay. There was no detectable UV absorbing material in the breakthrough and wash using a sensitivity of 0.1 on the monitor (Fig. 10). From the A280 reading of the fractions, the binding efficiency was greater than 90%. The endotoxin level of the affinity purified protein G was determined to be less than 50 E U mg-1. This level was achieved without any special precautions taken during the process.
Discussion
In recent years, membrane technology has emerged as an attractive alternative to the more traditional separation methods of downstream processing such as centrifugation and dead-end filtration because it offers an operation with a very
227 low labor requirement and no formation of aerosols. For CFM, permeate flux is a function of transmembrane pressure, mass transfer coefficient, temperature, tangential velocity, hydrodynamic conditions, membrane characteristics and solute/ particulate concentration (Mackay and Salusbury, 1988). Our data suggested that operating the cross-flow microfilter at excessive TMP decreased protein G transmission, probably due to the enhancement of concentration polarization and membrane fouling effects. At lower TMP, protein G transmission improved significantly with slight reduction of permeate flux. The flux achieved with lower TMP, 25-38 LMH, was slightly lower than the mean value suggested by the literature, 40-50 LMH (Kroner et al., 1984). However, Prostak operation, including solid-liquid separation and membrane clean up could easily be accomplished within an 8 h shift. A centrifugation process was also evaluated as a means of solid-liquid separation. A M-16 centrifuge (Sharpies) operated at 1 1 min -1 failed to remove the solids completely from the spent medium. A clear supernatant was obtained only after filtering the M-16 output through a Niagara filter (Ametek) with standard Supercel (Johns-Manville) as a filter aid. Centrifugation in combination with this polishing dead-end filtration process turned out to be more labor intensive than the CFM membrane process. When total protein was determined using the bicinchoninic acid assay, the range of purity (mg protein G per mg of total protein) in the Prostak output was calculated to be 0.02-0.17. When the Prostak output was dialyzed and concentrated using ultrafiltration, the purity increased significantly. Therefore, the majority of the impurities in the Prostak output were low molecular weight materials, such as amino acids, peptides and colored substances from the spent medium. This conclusion was supported by the SDS-PAGE analysis which also showed that there were only a few minor protein contaminants in the Prostak output (Fig. 8). A number of experimental breakthrough curves were generated from the laboratory affinity column in order to study the effect of changing operating variables on the adsorption efficiency of the pilot column. According to the breakthrough curves constructed, there were several means of optimizing large scale affinity column operation. In addition to the method employed herein, another option was to utilize the column to its full capacity and recycle a portion of the breakthrough material. This approach was demonstrated by Moks et al. (1987); in their report, the first 75% of the breakthrough containing less than 1% of product was discarded and the last 25% of the breakthrough was recycled. The affinity constant for streptococcal protein G binding to human IgG was reported to be 67.4 x 109 M -1 (/~kerstr6m and Bj6rck, 1986). For GammaBind ® G, it was determined to be 5.2 x 109 M -1 (Fahnestock, unpublished result). According to the above t w o g a values, the dissociation constant (K d) values were calculated to be 3.3 x 10 -7 and 4.2 x 10 -6 mg m1-1. Based on computer calculation, the shape and position of the breakthrough curve becomes constant when the adsorbate concentration is far greater than the K d value (Chase, 1984). The concentration of protein G loaded onto the column in these studies (0.18 and 1.1 mg m1-1) greatly exceeded the K~. This explained why we did not see any
228
significant difference in the breakthrough curve with different protein G concentrations. Comparing the breakthrough curves obtained with crude and purified protein G (Figs. 5, 6 and 7), crude protein G gave shallower breakthrough curves. This suggested that the media components, other proteins or possibly antifoam in the crude protein G preparation, were interfering with the interactions of protein G and IgG sepharose. An ultrafiltration system with a peristaltic pump has the advantages of low hold up volume and gentle pumping actions. However, the tubing is prone to fatigue failure and requires constant attention. Therefore, a Tri-flo ® centrifugal pump (Ladish Co., model Cl14) with SANI-PRO sanitary tubing (SaniTech) was installed into the Pellicon to replace the peristaltic pump. Although centrifugal pumps were known to denature long molecules with shearing, we had not experienced any shearing problems with protein G. A rotary lobe pump was recommended by Millipore, but it was much more expensive than a centrifugal pump.
Acknowledgements The authors acknowledge the assistance of Malcolm Finkelman, Marvin Stewart, Brian Bell, Bill Wilson, Lois Dinterman, Marie Wroble, David Frashier, Thomas MueUer, Phyllis Link and Jacki Ricks. We also thank Eric Rudolph and Burke Fahlman from Millipore for their assistance with the Prostak work.
References ,~kerstr6m, B. and Bj6rck, L. (1986) A physicochemical study of protein G, a molecule with unique immunoglobulin G binding properties. J. Biol. Chem. 261, 10240-10247. Bj6rck, L. and Kronvall, G. (1984) Purification and some properties of streptococcal protein G, a novel IgG binding reagent. J. Immunol. 133, 969-974. Boyle, M.D.P. and Reis, K.J. (1987) Bacterial Fc receptors. Bio/Technology 5, 697-703. Chase, H.A. (1984) Prediction of the performance of preparative affinity chromatography. J. Chromatogr. 297, 179-202. Fahnestock, S.R. (1987) Cloned streptococcal protein G genes. Trends Biotechnol. 5, 79-83. Fahnestock, S.R., Alexander, P., Nagle, J. and Filpula, D. (1986) Gene for an immunoglobulin-binding protein from a group G Streptococcus. J. Bacteriol. 167, 870-880. Finkelman, M.A.J, and Lee, T.K. (1990) Thermal release of recombinant protein into culture media. Int. Publ. WO 90/00200, International application published under the patent cooperation treaty, PCT/US89/01917. Kroner, K.H., Schiitte, H., Hustedt, H. and Kula, M.-R. (1984) Cross-flow filtration in the downstream processing of enzymes. Process Biochem. 19, 67-74. Laemmli, U.K. (1970) Cleavage of structural proteins during the assembly of the head of bacteriophage T4. Nature 227, 680-685. Layne, E. (1957) Spectrophotometric and turbidimetric methods for measuring proteins. In: Colowick, S.P. and Kaplan, N.O. (Eds.), Methods in Enzymology 3, Academic Press, New York, pp. 447-455. Mackay, D. and Salusbury, T. (1988) Choosing between centrifugation and cross-flow microfiltration. Chem. Eng. 447, 45-50.
229 Moks, T., Abrahmsen, L., Osterl6f, B., Josephson, S., Ostling, M., Enfors, S.-O., Persson, I., Nilsson, B. and Uhl6n, M. (1987) Large-scale affinity purification of human insulin-like growth factor I from culture medium of Escherichia coli. Bio/Technology 5, 379-382. Nakane, P.K. and Kawaoi, A. (1974) Peroxidase-labeled antibody, a new method of conjugation. J. Histochem. Cytochem. 22, 1084-1091. O'Sullivan, M.J. and Marks, V. (1981) Methods for the preparation of enzyme-antibody conjugates for use in enzyme immunoassay. In: Langone, J.J. and Van Vunakis, H. (Eds.), Methods in Enzymology 73, Academic Press, New York, pp. 147-166. Smith, P.K., Krohn, R.I., Hermanson, G.T., Mallia, A.K., Gartner, F.H., Provenzano, M.D., Fujimoto, E.K., Goeke, N.M., Olson, B.J. and Klenk, D.C. (1985) Measurement of protein using bicinchoninic acid. Anal. Biochem. 150, 76-85.
213
© 1992 Elsevier Science Publishers B.V. All rights reserved 0168-1656/92/$05.00
BIOTEC 00789
Process development for the recovery and purification of recombinant protein G S h w u - M a a n L e e , N i n a Z. Essig and T i m o t h y K. L e e Genex Corporation, 16020Industrial Drive, Gaithersburg, MD 20877, USA
(Received 22 April 1991;revisionaccepted 2 March 1992)
Summary The domains of protein G from streptococcus which bind immunoglobulin G have been cloned and expressed in Escherichia coli (Fahnestock et al., 1986). Because protein G binds to several animal immunoglobulin G's, it has many immunochemical applications. This report describes process development for large-scale production of this recombinant protein G (also known as GammaBind ® G). In 200 1 cultures of E. coli, this protein G variant was released from the cell into the culture medium by heating at 80°C for 10 min. The concentration was monitored by either a competitive enzyme-linked immunoassay or a liquid chromatographic assay. Cross-flow microfiltration with 0.22 Ixm membrane was used to remove the cells. The protein G-rich permeate from the crossfiow microfilter was purified by affinity chromatography using a 5 1 column of IgG-Sepharose 6 Fast Flow, which yielded 16-18 g of protein G per column cycle. The pools of purified protein G were concentrated and desalted using ultrafiltration. The salt-free protein G was then lyophilized as bulk product. The overall recovery through the entire process was 50-64%. The analysis of the final product included sodium dodecyl sulfate polyacrylamide gel electrophoresis, UV-visible spectrum, high performance gel filtration, endotoxin level and binding efficiency to human IgG Sepharose. Protein G; Cross-flow microfiltration; Downstream processing; Affinity chromatography; Ultrafiltration
Correspondence to: S.-M. Lee, AMVAX, Inc., 12040 Indian Creek Court, Beltsville, MD 20705, USA.
214 Introduction
Protein G is a cell wall protein of group G streptococci (Bj6rck and Kronvall, 1984; Boyle and Reis, 1987). Like protein A of Staphylococcus aureus, protein G binds to the Fc domains of a wide variety of immunoglobulin G's (IgG's). The potential applications of protein G include virtually all of the current and projected applications of protein A. These can be grouped into three categories: (a) the detection and quantitation of antibodies and immune complexes; (b) the purification of antibodies and immune complexes; (c) the removal of antibodies and immune complexes from serum of patients by extracorporeal plasma filtration. There has been considerable interest in using protein G in applications for which protein A is not well suited due to specificity differences. In general, protein G is more specific than protein A in that it binds only to IgG and not, for example, to IgM which protein A binds. On the other hand, protein G shows broader IgG reactivity, binding several important IgG's which protein A does not bind well. These include human IgG3, mouse IgGa, goat and rat IgG's. Until now, protein G has not been widely available because it is difficult to prepare from streptococci. The gene has been cloned from a streptococcus clinical isolate into Escherichia coli and Bacillus subtilis (Fahnestock et al., 1986; Fahnestock, 1987). The recombinant protein G (also known as GammaBind ® G) in this report refers to a modified form of protein G constructed by recombinant techniques and expressed in E. coli; for simplicity, the name protein G is used throughout this report. This recombinant protein G contains the two IgG binding B domains followed by a sequence rich in proline residues and a set of five 'C-repeats', analogs of the pentapeptide Asp-Asp-Ala-Lys-Lys, near the C-terminus (Fahnestock, 1987). The deleted sequences include those which are responsible for albumin binding in protein G. This report describes the assay development and process development for large scale production of this recombinant protein G.
Materials and Methods
Competitive enzyme-linked immunosorbent assay (ELISA) for protein G determination Protein G-horseradish peroxidase (HRP) conjugate was prepared according to the periodate-oxidation method (O'Sullivan and Marks, 1981; Nakane and Kawaoi, 1974). The weight ratio of HRP to protein G used in conjugation was 20 to 1; the conjugate mixture was not separated from unconjugated proteins. Microtiter plates (Immulon 2, Dynatech) were coated at room temperature overnight with 110 ~xl (110 ng) of human IgG (Jackson Immunoresearch Lab) in 50 mM sodium phosphate buffer containing 0.15 M NaC1, pH 7.4. The coating solution was removed and the plates were blocked by incubation for 1 h or longer at room temperature with 200 Ixl of casein (Fisher), which had been prepared as a 2% solution in 7 mM sodium phosphate buffer, pH 7.4, containing 0.1 M NaCI and 0.2% sodium azide. Before sample and standard were added, the plates were
215 washed once with 10 mM sodium phosphate buffer, 0.15 M NaC1, pH 7.4, containing 0.05% Tween 20 (Sigma). Protein G standard was prepared by affinity purification through IgG Sepharose 6 Fast Flow and its concentration determined using an extinction coefficient of 1.0 for a 1 mg m1-1 solution at 280 nm. The standard and samples were prediluted in the tubes, followed by a twofold serial dilution in the plates and incubated for 1 h. To the plates containing 100 ixl of sample or standard, 10 txl of protein G - H R P was added and the incubation proceeded for 30 min. The conjugate was prepared by diluting a stock solution of the conjugate (protein G concentration at 0.06 mg m l - l ) 5000-fold with 50 mM sodium phosphate buffer, 0.15 M NaC1, pH 7.4, containing 0.1% bovine serum albumin (Sigma, RIA grade) and 0.05% Tween 20. The same dilution buffer was used for both standard and samples. The plates were washed ten times after the conjugate incubation and the color was developed with the 3,3',5,5'-tetramethyl benzidine substrate system (Kirkegaard and Perry Lab). The color reaction was stopped at 30 min with 1 M phosphoric acid. The absorbance at 450 nm was measured with a multiscan spectrophotometer (Titertek).
Liquid chromatographic assay for protein G determination Liquid chromatography (LC) was performed on a Waters liquid chromatography system consisting of a model 720 controller, a model 730 data module, a model 6000A pump, a M-45 pump, a model 440 absorbance detector, a U6K injector with a 2 ml loop and an analytical column of immobilized human IgG. Crude protein G sample or affinity purified protein G standard at pH 6.5-7.5 was loaded at a flow rate of 1 ml min-1 onto the column which had been equilibrated with 45 mM potassium phosphate buffer containing 150 mM NaC1 and 0.005% (w/v) sodium azide, pH 7.4 (PBS/NaN3). The column was then washed with the same buffer to remove unadsorbed material at a flow rate of 5 ml min-1. The adsorbed protein G was eluted with 0.5 M ammonium acetate, pH 3, at a flow rate of 5 ml min-1. Following elution, the column was re-equilibrated with P B S / N a N 3. The run time was 20 min. From the chromatogram, the peak height was matched to a standard curve to determine the amount of protein G present in the sample.
Recovery of protein G from spent medium At the end of an E. coli cultivation, protein G was released from the cells into the spent medium by heating the bioreactor at 80°C for 10 min (Finkelman and Lee, 1990). During the early phase of process development, the spent medium was often aged at 5°C for 1-3 days after the heating process to facilitate the release of protein G from the cells. The cells were separated from the protein G-containing spent medium by cross-flow microfiltration (CFM) through a Prostak-1 pilot unit (Millipore) using a 4.6 m 2 membrane (GVPP type, 0.2 ~m, wide channel, Millipore) operated at ambient temperature. Initially the system was drained and the integrity of the membrane was tested in the system by measuring air diffusion rate at 2 bar to be less than 14.5 ml m -2 min -1. Once the integrity of the membrane was confirmed, the cells in the spent medium were concentrated 10-20-fold across the membrane
216
and the protein G-rich permeate was collected. The concentrated cells (retentate) were then washed continuously with 100-200 1 of P B S / N a N 3 on the Prostak membrane. At the end of CFM, the retentate was discarded and the system was cleaned in accordance with the manufacturer's instruction which included saline wash, water wash, detergent wash, sodium hypochlorite sanitization and warm acetic acid wash. Following a water wash, the system was stored in 0.05% sodium azide.
Purification of protein G IgG Sepharose 6 Fast Flow (Pharmacia) was suspended in 20 mM Tris buffer, pH 7.5, and packed in a 5-1 column (25.2 cm × 10 cm, Pharmacia) at 5°C. The flow was maintained with a peristaltic pump (Watson-Marlow type 604 U / R , Bacon Technical Industries, Inc.). Two on-line filters (1 and 5 txm Polycap TM 75HD, Whatman) were used to protect the column from particulates and a pressure gauge was installed between the pump and the filters. The column was initially washed with 20 mM Tris buffer, pH 7.5, to remove MgClz and ethanol (preservatives) and subsequently washed with 1 M ammonium acetate, pH 3, to remove unbound lgG from the gel matrix. P B S / N a N 3 was prepared as a fivefold concentrate and clarified by passing through a 1/xm polypropylene cartridge filter (Pall). The column was equilibrated with PBS/NaN3, the output from Prostak was loaded onto the column in an upward direction using a linear flow rate of 150 cm h-1 (75 I h-1), which produced an operating pressure of 1 bar. The column was washed with P B S / N a N 3 (75-150 1) until the absorbance at 280 nm dropped to baseline as indicated by the on-line UV monitor (Type UA-5, ISCO). The column was subsequently washed with 5 mM ammonium acetate buffer, pH 5.0, (25-50 1) to remove the bulk of the salt from the column and bring the pH down to 5. The column was eluted with 0.5 M ammonium acetate, pH 3, in a downward direction using a linear flow rate of 75 cm h-1. This protein G solution was adjusted to pH 5 with concentrated ammonium hydroxide and stored at 5°C. The column was equilibrated again with P B S / N a N 3 and stored in the same buffer at 5°C.
Ultrafiltration The purified protein G was pooled for concentration and dialysis. A Pellicon system (Millipore) with a 0.46 m 2 polysulfone membrane cassette (PTGC type, 10000 nominal molecular weight cut off, Millipore) was used. The flow was controlled by a peristaltic pump (Watson-Marlow type 604 U / R ) . The membrane cassette was integrity tested at 0.3 bar inlet pressure and confirmed that the air diffusion rate was less than 10 ml rain-1. The ultrafiltration system was operated at a recirculation rate of 4 1 min-1 at room temperature. Following the initial concentration, further dialysis was conducted by batch addition of equal volumes of 2 mM N H a H C O 3 buffer, pH 7. The desalting process was monitored by measuring the conductivity of the permeate. The system was cleaned by recirculation of 0.1 N NaOH. Following a water wash to remove the residual NaOH, the system was stored in 0.05% NaN 3.
217
Lyophilization The salt-free protein G solution was clarified by 0.22 I~m filtration (Millipak 60, Millipore) and lyophilized in bulk using a Freeze Mobile 24 (Virtis).
Sodium dodecyl sulfate polyacrylamide gel electrophoresis (SDS-PAGE) SDS-PAGE was performed according to the method of Laemmli (1970). Gels of 1.5 mm thickness were constituted with 15% polyacrylamide and stained with Coomassie blue. Molecular weight protein markers (Bio-Rad) included phosphorylase b, 97 kDa; bovine serum albumin, 67 kDa; ovalbumin, 43 KDa; carbonic anhydrase, 31 kDa; soybean trypsin inhibitor, 22 kDa; and lysozyme, 14 kDa.
Protein determination The concentration of affinity purified protein G was determined by its UV absorbance using an extinction coefficient of 1.0 at 280 nm for a 1 mg m1-1 (0.1%) solution. The protein concentration in the crude protein G preparation was determined using the Pierce bicinchoninic acid assay (Smith et al., 1985). Affinity purified protein G was used as a standard to construct the calibration curve.
HPLC gel filtration assay for purity determination We used a TSK G2000SWXL column (0.78 cm x 30 cm, TosoHaas) with its guard column (0.6 cm x 4 cm). The running and column storage buffer was 0.1 M sodium phosphate containing 0.1 M sodium sulfate and 0.05% sodium azide, pH 6.4. The flow rate used was 0.5 ml min-1 and the total run time was 30 min.
Binding efficiency assay An IgG Sepharose 6 Fast Flow column (1.6 cm X 3 cm) was pre-eluted with 1 M acetic acid and equilibrated with P B S / N a N 3. The flow rate was maintained at 0.6 ml min-1 with a peristaltic pump. Lyophilized, salt free protein G was dissolved in P B S / N a N 3 to a final concentration of 1.2 mg m1-1 and a total of 5 ml was applied to the column. Following the column wash with P B S / N a N s , the protein G bound to the column was eluted with 1 M acetic acid. The 280 nm absorbance of each individual fraction was measured using a spectrophotometer (Shimadzu, UV-260). The binding efficiency was calculated according to the following formula: A280 eluted Binding efficiency (%) = A2s0 in breakthrough and wash + A2s 0 eluted
Endotoxin assay Endotoxin was determined using the gel-clot method performed by Associates of Cape Cod, Inc. Woods Hole, MA. A total of 9 mg of lyophilized protein G was reconstituted with 1 ml of endotoxin free water and titered down using a twofold dilution scheme. E. coli endotoxin standard was added to sample dilutions as inhibition controls. The assay sensitivity was 0.03 E U ml-1.
218
Results Protein G assay
In the initial process development stages, the competitive ELISA was used to monitor protein G concentration in the crude sample. In this assay, protein G competed with protein G - H R P conjugate for binding to human IgG adsorbed onto microtiter plates. The working range was 30-1000 ng m1-1 (Fig. 1). The assay was adequate to determine protein G concentration in the spent medium following heat release. However, the major drawback was its speed as it usually required 4 h to complete one assay. This assay was therefore impractical to use for in process monitoring. The specific binding between protein G and human IgG as well as the availability of an immobilized IgG analytical column allowed us to establish an LC assay. Figure 2 shows a typical LC chromatogram. The assay was validated by adding a protein G standard to a typical crude protein G sample (i.e., the Prostak output) from which protein G had been specifically removed by IgG Sepharose 6 Fast Flow. The standard curve remained the same in the presence of the contaminants in the Prostak output (Fig. 3). It was concluded that there was minimal interference from the contaminants and that the eluted peak in the LC assay indeed represented the amount of protein G in the crude sample. The capacity of the immobilized IgG analytical column was determined to be 200 Ixg. A convenient working range was 20-160 I~g of protein G per assay. With the 2 ml injection loop the detection limit was 10 Fxg m l - 1. This assay was adequate to support the process development work. The immobilized IgG analytical column had an average life of six months or 300-500 injections under our operating conditions. As the column deteriorated,
2.
[ o I/1
W C in I. O
0
. . . . . . . .
1o
f
1oo
. . . . . . . .
i
1ooo
. . . . . . . .
1oooo
Protein G (ng ml -1) Fig. 1. R e s p o n s e of c o m p e t i t i v e E L I S A to various c o n c e n t r a t i o n s of p r o t e i n G s t a n d a r d . T h e points on the curve r e p r e s e n t d u p l i c a t e v a l u e s p e r f o r m e d on the s a m e plate.
219
IA(280nm) Loacl
Wash
[
Elutlon
I
Reequlllbratlon
I
110
1
I
15
Time (mln) Fig. 2. Immobilized IgG analytical column loading and elution profile. A total of 300 ~1 of standard protein G at a concentration of 0.48 mg m l - 1 was manually injected onto the column. The flow rate was 5 ml m i n - 1 except during the loading stage when it was 1 ml min 1. The full scale absorbance was 0.2 at 280 nm.
the peak height dropped and the assay sensitivity was diminished. The advantages of this assay included short analysis time (20 min), relative simplicity and reproducibility compared to the ELISA.
15
10 v u
~t
5
Q,.
0 0
100
200
Protein G (/Jg) Fig. 3. Standard curve obtained by injecting known amounts of protein G onto an immobilized IgG analytical column. Affinity purified protein G solution at a concentration of 0.48 mg m l - I was applied to the column at volumes of 100-400 ixl (e). The same standard was diluted 5-fold with Prostak output from which the protein G had been specifically removed by IgG Sepharose 6 Fast Flow and 0.5-2 ml of the diluted standard was loaded onto the column (o). Elution peak height was plotted as a function of the a m o u n t of protein G loaded.
220
Solid-liquid separation The solids content in the protein G-released spent medium was usually about 2.5% as measured in a Gyrotester (Alfa-laval). This relatively low solids content made CFM an attractive approach for initial solid-liquid separation. Initially, the Prostak unit was operated at a cross flow rate of 170-190 1 min-1 which produced an inlet pressure of 3.4-3.9 bar and a retentate pressure of 0.8-1.0 bar. The p e r m e a t e and retentate valves were fully opened. The average transmembrane pressure (TMP) was calculated to be 2.1-2.5 bar according to this formula: T M P = (inlet pressure + retentate pressure) - 2. Using these operating parameters, the p e r m e a t e flow rate was measured to be 2.8-3.5 1 m i n - 1 which was equivalent to 37-46 1 m -2 h -x (LMH). The percentage of protein G transmitted through the m e m b r a n e was 39% initially and gradually declined to 20% when the cells were concentrated sixfold. The protein G recovered in the initial 10-fold concentration of the cells was less than 50%. In combination with 180 1 of cell washing, the yield was about 75%. The operating parameters had to be modified to improve the transmission and yield. The cross flow rate was reduced to 135 1 min -~ which produced an inlet pressure of 2.6-2.9 bar. The retentate gauge indicated 0.7-0.9 bar of pressure with the valve fully opened. A diaphragm valve was installed in the p e r m e a t e stream to adjust TMP, which was calculated according to the following formula: T M P = (inlet pressure + retentate pressure) + 2 - permeate pressure A p e r m e a t e pressure of 0.7 bar was used which produced a T M P of 1.0-1.2 bar. Figure 4 shows the transmission and p e r m e a t e flux along the primary concentration process using the modified procedure with 1.0-1.2 bar TMP. When compared to the original operating parameters with higher TMP, about twice as much protein G transmitted through the m e m b r a n e with only a slight reduction of p e r m e a t e flux. Table 1 shows the result of a typical run; 72% of the input protein G was recovered in the p e r m e a t e through the initial cell concentration step and 28% through the washing step. The final retentate contained only 2.7% of the input protein G and the yield was quantitative when sufficient wash was used. Without the washing step, we expected a 72% yield.
Affinity chromatography In the m o d e of packed column operation, the sample is fed continuously into the column until the available capacity of the column (IgG Sepharose 6 Fast Flow) is exhausted and the adsorbate (protein G) begins to appear in the column effluent. The variation of the concentration of adsorbate in the column effluent as a function of the amount of adsorbate applied to the column is a breakthrough curve (Chase, 1984). To understand the interaction between protein G and I g G Sepharose and to optimize the affinity process, breakthrough curves were established in a laboratory column (1.6 cm x 10 cm) which was kept at the same length as the pilot column (25.2 cm × 10 cm). The cross-sectional areas for the pilot and laboratory columns
221
50
100
A
90"
~
Transmission • 40
A "r"
80'
--1 v • 30
E ¢n
70@ • 20
60
E II.
10 a.
50
0
40
10
Concentration
factor
Fig. 4. Protein G transmission and permeate flux during the course of cell separation in the Prostak under 1.0-1.2 bar TMP. Concentration factor is the ratio of original volume to retentate volume at various points along the cell concentration process. Protein G transmission is the ratio of protein G concentration in the permeate to protein G concentration in the retentate. The LC assay was used for protein G determination.
were 500 and 2 cm 2, respectively. Therefore, the scale-up factor in terms of column bed volume and column capacity was the ratio of the two cross-sectional areas, i.e., 250. The Prostak permeate with protein G concentration adjusted to 0.18 mg m1-1 was continuously fed into the IgG Sepharose column (1.6 cm x 10 cm) until the available capacity of the column was totally exhausted. A linear flow rate of 150 cm h-1 was used. The breakthrough curve was constructed by plotting the percentage of protein G appearing in the breakthrough fractions against the amount of protein G loaded (Fig. 5). The yield curve was established by plotting the
TABLE 1 Protein G recovery through Prostak pilot-1 unit
Protein G-containing spent m e d i u m Permeate collected through initial cell concentration Permeate collected during cell washing stage Final retentate T M P of 1.0-1.2 bar was used.
Normalized protein G concentration
Volume (1)
Yield (%)
1.0 0.76
190 180
100 72
0.28
185
28
0.25
20
2.7
222
100
o~"
• 100
80
' 80
40
.40
o
.S
~.
o 20
rough
/ 0
n.~ 50
,20
Yield 0 100
1 50
200
Amount of protein G loaded (rag)
Fig. 5. Breakthrough curve and yield curve at a linear flow rate of 150 cm h -1. Prostak output with protein G concentration adjusted to 0.18 mg m1-1 was loaded onto an IgG Sepharose 6 Fast Flow column (1.6 cm × 10 cm) at 5 ml m i n - 1. The LC assay was used to determine protein G concentration in the breakthrough fractions and the column starting material. Protein G eluted from the column was determined by UV absorbance at 280 nm.
percentage of protein G recovered against the different amounts of protein G loaded (Fig. 5). The column's saturation capacity was determined to be 4 mg of protein G per ml of gel. The laboratory column's full capacity was 80 mg. The breakthrough curve in Fig. 5 indicated that when 56 mg of protein G was applied to the coltrmn, protein G started to leak from the column and the percentage leaking increased as the loading amount increased. Beyond this breakthrough point, the more protein G loaded, the more protein G would be lost in the breakthrough and the lower the yield. In a production situation, it would be the best to utilize an affinity column's full capacity so as to decrease the number of production cycles. However, to completely saturate the column, more than 160 mg of protein G (twice the column capacity) had to be applied and the yield would be lower than 50%. Therefore, if the IgG Sepharose column was to be fully utilized with a reasonable yield, the protein G lost in the breakthrough had to be recycled. Applying 70-90 mg of crude protein G to the laboratory column, which was equivalent to 18-23 g to the pilot column, appeared to be a better compromise for routine operation; the yield under these conditions was predicted to be 65-75%. With this loading amount, there was a small loss in the breakthrough, but most of the column capacity was utilized. In an affinity chromatography system, the most efficient adsorption performance is obtained when the deflection point of the breakthrough curve is the sharpest. The concentration of adsorbate and loading flow rate can both affect the shape of the breakthrough curve (Chase, 1984). The effect of loading flow rate was
223
80
80 A
-1
v .C
6O
a
-
150 cm h
•
75 cm
,60
1
4O
.40
2O
.20
.E 0
i
O.
50
100
150
200
Amount of protein G loaded (rag) Fig. 6. Breakthrough curves at linear flow rates of 150 and 75 cm h - I. The starting material for the IgG Sepharose column (1.6 cm × 10 cm) was affinity purified protein G at concentration of 0.33 mg m l - 1 . T h e protein G concentration in the breakthrough was determined by U V absorbance at 280 nm.
assessed by loading purified protein G onto the laboratory column at 150 and 75 cm h - t (Fig. 6). Applying 70-90 mg of protein G to the laboratory column with a saturation capacity of 80 mg, one would expect a slightly higher yield with a slower flow rate as indicated in the breakthrough curves. However, the slight improvement in yield with the slower flow rate did not justify spending additional production time in loading. The effect of adsorbate concentration on the shape of the breakthrough curve was also investigated. For this experiment, the Prostak permeate was adjusted to a final protein G concentration of 1.1 mg ml-1. When this material was compared to that of 0.18 mg ml-1, the two breakthrough curves overlapped (Fig. 7). Although, concentrating the Prostak output reduced the loading time, there was no improvement in the adsorption efficiency and yield. Table 2 presents data from the pilot affinity column. As predicted from the development work, the first half of the breakthrough contained no protein G and the second half contained a small amount. In this example, 83% of the column capacity was utilized and the yield was 72%.
Ultrafiltration The ultrafiltration process was optimized by conducting a pressure excursion experiment. The permeate flow rate was measured at 0.14 bar increments of average TMP [(inlet pressure + retentate p r e s s u r e ) - 2]. The optimized average TMP was determined to be 1.3 bar, which was the point where increasing TMP did not increase the flux. The permeate flow rate achieved initially was 0.26 1 min-1 and gradually slowed down to 0.15 1 min -1. Protein G was concentrated 28-fold with 87% recovery, to give a solution of 46.2 g protein in 2.1 1.
224
A
80
- 80
60
-60
40
"40
20
"20
=,..
o
.,Q
0 Q.
•
20
40
60
80
100
!
120
140
Amount of protein G loaded (rag) Fig. 7. Breakthrough curves established at a linear flow rate of 150 cm h - i using two different starting materials with protein G concentrations of 0.18 and 1.1 mg ml - I . Protein G was determined by LC assay.
Overall recovery The yield for initial solid-liquid separation was 90-98% depending on the amount of wash used. The yield for affinity chromatography was 65-75% depending on the amount of protein G loaded. The yield through ultrafiltration step was generally 85-87%. Therefore the total yield throughout the entire process was 50-64%.
Analysis of affinity purified protein G Affinity purified protein G was characterized by its absorbance maximum at 278 nm, absorbance minimum at 250 nm and a distinct shoulder at 283 nm. The A280/A260 ratio was calculated to be 1.77-2.00 in several lots tested. According to Layne's method (1957), an A280/m26 o ratio of lower than 1.75 indicates significant nucleic acid contamination of a protein solution. Therefore, the affinity purified protein G was low in nucleic acids.
TABLE 2 Protein G recovery from pilot affinity column Total protein G
Yield (%)
(g) Starting material (Prostak output) Breakthrough fraction 1 Breakthrough fraction 2 Column wash Elution
23.0 0 1.9 1.3 16.6
100 0 8 6 72
The bed volume of the column was 5 1 (25.2 cm × 10 cm) and flow rate was set at 150 cm h - i (75 1 h - 1).
225
1
2
3
4 ---97 K "-'67 K ---43K ---31 K
---22K --14 K
Fig. 8. SDS-PAGE of crude and purified protein G. Lane 1: Prostak output containing 20 txg of protein G; Lanes 2 and 3: affinity purified protein G, 20 and 1 I~g, respectively; Lane 4: molecular weight standards. F i g u r e 8 s h o w s t h e S D S - P A G E o f c r u d e a n d p u r i f i e d p r o t e i n G. S i n c e p r o t e i n G w a s r e l e a s e d f r o m t h e cells w i t h o u t m e c h a n i c a l b r e a k a g e , t h e l e v e l o f p r o t e i n c o n t a m i n a n t s in t h e s p e n t m e d i u m was v e r y low. T h e p r e d o m i n a n t c o m p o n e n t in
~o
vI d
A
Time (min) Fig. 9. HPLC gel filtration analysis of affinity purified protein O. Lyophilized protein G was reconstituted in 0.1 M sodium phosphate, pH 6.4, containing 0.1 M sodium sulfate and 0.05% sodium azide to a final concentration of 1.4 mg ml- 1. A total of 15 p,l was injected and the elution was monitored at 280 nm. The full scale absorbance was 0.1. V0 indicates the void volume determined by blue dextran and Vi is the included volume determined by p-aminophenylalanine.
226
0.05
Load
Wash
0.5 I
Elution
A (280 nm)
I
2 T i m e (h)
Fig. 10. U V monitor tracing from binding efficiency assay. The full scale absorbance was set at 0.1 for loading and washing stages and set at 1.0 for elution. The baseline was readjusted upon switching to 1.0
absorbance. the Prostak output was protein G with only a few minor impurities (lane 1). Protein G purified from the current procedure appeared as essentially a single major band on the SDS gel (lanes 2 and 3). SDS-PAGE was a sensitive way to detect contaminating proteins, and had been extremely useful in monitoring our process development work. However, it was difficult to quantitate protein G purity reproducibly based on scanning of stained gel. The HPLC gel filtration assay offered a more quantitative method for determining protein G purity. As shown in Fig. 9, two minor impurities were detected using this method. The purity was calculated to be 97% using integration of the peak areas. The affinity purified protein G was reapplied to the IgG Sepharose Fast Flow column as described in the binding efficiency assay. There was no detectable UV absorbing material in the breakthrough and wash using a sensitivity of 0.1 on the monitor (Fig. 10). From the A280 reading of the fractions, the binding efficiency was greater than 90%. The endotoxin level of the affinity purified protein G was determined to be less than 50 E U mg-1. This level was achieved without any special precautions taken during the process.
Discussion
In recent years, membrane technology has emerged as an attractive alternative to the more traditional separation methods of downstream processing such as centrifugation and dead-end filtration because it offers an operation with a very
227 low labor requirement and no formation of aerosols. For CFM, permeate flux is a function of transmembrane pressure, mass transfer coefficient, temperature, tangential velocity, hydrodynamic conditions, membrane characteristics and solute/ particulate concentration (Mackay and Salusbury, 1988). Our data suggested that operating the cross-flow microfilter at excessive TMP decreased protein G transmission, probably due to the enhancement of concentration polarization and membrane fouling effects. At lower TMP, protein G transmission improved significantly with slight reduction of permeate flux. The flux achieved with lower TMP, 25-38 LMH, was slightly lower than the mean value suggested by the literature, 40-50 LMH (Kroner et al., 1984). However, Prostak operation, including solid-liquid separation and membrane clean up could easily be accomplished within an 8 h shift. A centrifugation process was also evaluated as a means of solid-liquid separation. A M-16 centrifuge (Sharpies) operated at 1 1 min -1 failed to remove the solids completely from the spent medium. A clear supernatant was obtained only after filtering the M-16 output through a Niagara filter (Ametek) with standard Supercel (Johns-Manville) as a filter aid. Centrifugation in combination with this polishing dead-end filtration process turned out to be more labor intensive than the CFM membrane process. When total protein was determined using the bicinchoninic acid assay, the range of purity (mg protein G per mg of total protein) in the Prostak output was calculated to be 0.02-0.17. When the Prostak output was dialyzed and concentrated using ultrafiltration, the purity increased significantly. Therefore, the majority of the impurities in the Prostak output were low molecular weight materials, such as amino acids, peptides and colored substances from the spent medium. This conclusion was supported by the SDS-PAGE analysis which also showed that there were only a few minor protein contaminants in the Prostak output (Fig. 8). A number of experimental breakthrough curves were generated from the laboratory affinity column in order to study the effect of changing operating variables on the adsorption efficiency of the pilot column. According to the breakthrough curves constructed, there were several means of optimizing large scale affinity column operation. In addition to the method employed herein, another option was to utilize the column to its full capacity and recycle a portion of the breakthrough material. This approach was demonstrated by Moks et al. (1987); in their report, the first 75% of the breakthrough containing less than 1% of product was discarded and the last 25% of the breakthrough was recycled. The affinity constant for streptococcal protein G binding to human IgG was reported to be 67.4 x 109 M -1 (/~kerstr6m and Bj6rck, 1986). For GammaBind ® G, it was determined to be 5.2 x 109 M -1 (Fahnestock, unpublished result). According to the above t w o g a values, the dissociation constant (K d) values were calculated to be 3.3 x 10 -7 and 4.2 x 10 -6 mg m1-1. Based on computer calculation, the shape and position of the breakthrough curve becomes constant when the adsorbate concentration is far greater than the K d value (Chase, 1984). The concentration of protein G loaded onto the column in these studies (0.18 and 1.1 mg m1-1) greatly exceeded the K~. This explained why we did not see any
228
significant difference in the breakthrough curve with different protein G concentrations. Comparing the breakthrough curves obtained with crude and purified protein G (Figs. 5, 6 and 7), crude protein G gave shallower breakthrough curves. This suggested that the media components, other proteins or possibly antifoam in the crude protein G preparation, were interfering with the interactions of protein G and IgG sepharose. An ultrafiltration system with a peristaltic pump has the advantages of low hold up volume and gentle pumping actions. However, the tubing is prone to fatigue failure and requires constant attention. Therefore, a Tri-flo ® centrifugal pump (Ladish Co., model Cl14) with SANI-PRO sanitary tubing (SaniTech) was installed into the Pellicon to replace the peristaltic pump. Although centrifugal pumps were known to denature long molecules with shearing, we had not experienced any shearing problems with protein G. A rotary lobe pump was recommended by Millipore, but it was much more expensive than a centrifugal pump.
Acknowledgements The authors acknowledge the assistance of Malcolm Finkelman, Marvin Stewart, Brian Bell, Bill Wilson, Lois Dinterman, Marie Wroble, David Frashier, Thomas MueUer, Phyllis Link and Jacki Ricks. We also thank Eric Rudolph and Burke Fahlman from Millipore for their assistance with the Prostak work.
References ,~kerstr6m, B. and Bj6rck, L. (1986) A physicochemical study of protein G, a molecule with unique immunoglobulin G binding properties. J. Biol. Chem. 261, 10240-10247. Bj6rck, L. and Kronvall, G. (1984) Purification and some properties of streptococcal protein G, a novel IgG binding reagent. J. Immunol. 133, 969-974. Boyle, M.D.P. and Reis, K.J. (1987) Bacterial Fc receptors. Bio/Technology 5, 697-703. Chase, H.A. (1984) Prediction of the performance of preparative affinity chromatography. J. Chromatogr. 297, 179-202. Fahnestock, S.R. (1987) Cloned streptococcal protein G genes. Trends Biotechnol. 5, 79-83. Fahnestock, S.R., Alexander, P., Nagle, J. and Filpula, D. (1986) Gene for an immunoglobulin-binding protein from a group G Streptococcus. J. Bacteriol. 167, 870-880. Finkelman, M.A.J, and Lee, T.K. (1990) Thermal release of recombinant protein into culture media. Int. Publ. WO 90/00200, International application published under the patent cooperation treaty, PCT/US89/01917. Kroner, K.H., Schiitte, H., Hustedt, H. and Kula, M.-R. (1984) Cross-flow filtration in the downstream processing of enzymes. Process Biochem. 19, 67-74. Laemmli, U.K. (1970) Cleavage of structural proteins during the assembly of the head of bacteriophage T4. Nature 227, 680-685. Layne, E. (1957) Spectrophotometric and turbidimetric methods for measuring proteins. In: Colowick, S.P. and Kaplan, N.O. (Eds.), Methods in Enzymology 3, Academic Press, New York, pp. 447-455. Mackay, D. and Salusbury, T. (1988) Choosing between centrifugation and cross-flow microfiltration. Chem. Eng. 447, 45-50.
229 Moks, T., Abrahmsen, L., Osterl6f, B., Josephson, S., Ostling, M., Enfors, S.-O., Persson, I., Nilsson, B. and Uhl6n, M. (1987) Large-scale affinity purification of human insulin-like growth factor I from culture medium of Escherichia coli. Bio/Technology 5, 379-382. Nakane, P.K. and Kawaoi, A. (1974) Peroxidase-labeled antibody, a new method of conjugation. J. Histochem. Cytochem. 22, 1084-1091. O'Sullivan, M.J. and Marks, V. (1981) Methods for the preparation of enzyme-antibody conjugates for use in enzyme immunoassay. In: Langone, J.J. and Van Vunakis, H. (Eds.), Methods in Enzymology 73, Academic Press, New York, pp. 147-166. Smith, P.K., Krohn, R.I., Hermanson, G.T., Mallia, A.K., Gartner, F.H., Provenzano, M.D., Fujimoto, E.K., Goeke, N.M., Olson, B.J. and Klenk, D.C. (1985) Measurement of protein using bicinchoninic acid. Anal. Biochem. 150, 76-85.
