Development of a Novel Platform TFF System for Insect Cell Culture Harvest

June 2, 2006
James Kacmar

James Kacmar, PhD is a process development scientist at NCRST, Inc.

Hank Kopf

Hank Kopf is chief technology officer at NCRST, Inc.

Richard Chubet

Richard Chubet is a scientist at Protein Sciences Corporation

BioPharm International, BioPharm International-06-02-2006, Volume 2006 Supplement, Issue 3

The overall average flux rate for the concentration and diafiltration step was 41 L m–2 h–1.

Severe Acute Respiratory Syndrome (SARS), a coronavirus (SARS CoV) is a respiratory disease, the main symptoms of which include fever, cough, shortness of breath, and pneumonia. Because SARS has the potential to reappear as a new, naturally acquired outbreak or by accidental or intentional release, an effective vaccine is sought. One of the most promising candidates for a SARS vaccine is the S glycoprotein protein. This glycoprotein forms large spikes in the viral envelope and mediates the binding of SARS CoV to the host cell through the host cell receptor, Angiotensin Converting Enzyme II (ACE2).1 A promising platform for the production of this SARS vaccine is insect cell cultures. Insect cells have the ability to produce proteins at high concentrations with translational modifications that permit proper folding and function. Additional advantages of insect cells over mammalian cells include the ease of culture, high molarity tolerance, and high expression levels.

This increased demand for insect cell products leads, in turn, to greater need for effective harvesting methods for large-scale cultures. Techniques have been established for harvesting insect cells using batch centrifugation,2 continuous centrifugation,3 and tangential flow filtration using either hollow fibers4 or flat sheets.5 Batch centrifugation is the simplest method for small-scale cultures, but it is very difficult to scale up to larger sizes because each batch usually is limited to only a few liters. Continuous centrifugation permits larger scale operation, but requires considerable initial investment. Also, the concentrate contains many particles that are 1 µm or larger, which means an additional filtration step is necessary before further processing. Finally, the product that remains in the sediment-containing phase is not recovered.6

An alternative to centrifugation is tangential flow filtration (TFF). TFF has the advantage of being linearly scalable from the bench top to large scale production.7 A major drawback to TFF is the tendency of microfiltration membranes to foul with particles in the media, which lowers membrane performance and makes cleaning difficult.8,9 Another complication of using TFF is the shear sensitivity of infected insect cells.5 If TFF is operated at too high a flow rate, the cells will break apart as a result of the shear. These cell parts further foul the membrane.

This study tests the performance of a modified TFF system, "SmartFlow TFF" (SFTFF, NCSRT, Apex, NC), in obtaining a high yield of a target protein from insect cell culture. In SFTFF, ribs create uniform retentate channels with uniform flow patterns over the entire functional surface of the membrane module. These uniform flow patterns maximize membrane efficiency and minimize fouling of the membrane module. The uniform flow also makes it possible to scale the system linearly from the laboratory bench to commercial production. In installed systems for other applications, the system has been scaled up directly from 5-m2 laboratory studies to commercial operations that process process 1.4 million L/d.

In this work, a single tank, single-module process was developed using SFTFF to harvest the SARS-CoV spike protein, which was expressed intracellularly from an insect cell culture. One microfiltration module was used for the following steps: (i) clarification; (ii) diafiltration to replace the medium with buffer; (iii) lysis of cells and release of the desired protein by the addition of extraction buffer; and (iv) passage of the released protein through the membrane. Then, an ultrafiltration membrane was used to concentrate the protein and perform a buffer exchange before loading the protein onto an HPLC column. No lysis was observed, which means that SFTFF did minimal harm to the cells during clarification. The one tank–one module method simplifies the harvest in comparison to multiple centrifugation steps. The implications of these results for both small-scale and large-scale harvesting of insect cells are discussed below.

Materials and Methods

Cell Growth Conditions

The gene encoding the S glycoprotein was cloned from a lysate of culture SARS-CoV 3200300841 (Passage #3) in Trizol LS Reagent (Sigma, St. Louis, MO), kindly provided by Dr. Dean Erdman (CDC, Atlanta, GA). The cDNA was cloned into a baculovirus transfer vector pSC12 (Protein Sciences Corporation, Meriden, CT) downstream from the baculovirus very late promoter of the polyhedron gene and in frame with the chitinase secretion signal sequence at the N-terminus and 6-his at the C terminus.

To generate a recombinant baculovirus, linearized Autographa California Nuclear Polyhedrosis Virus (AcNPV) DNA, and the recombination plasmid DNA containing the S gene were mixed and co-precipitated with calcium phosphate, and Sf9 cells were transfected. Recombinant viruses were identified by their distinctive plaque morphology. A single recombinant virus plaque was isolated. The virus was further amplified using Sf9 cells for P1 and ExpresSF+ cells in the absence of serum for P2 and P3. The recombinant virus was further scaled up before added to 10 L of ExpresSF+ cell culture in a bioreactor at a multiplicity of infection of 1.0 plaque forming unit/cell. The infected culture was incubated at 27–308 C for 48 h. During this period, recombinant SARS-CoV FL–His-6 was not secreted and was associated intracellularly.

Filtration Setup and Operation

A PUROSEP LT-2Q filtration skid equipped with an OPTISEP 3000 holder (NCSRT, Inc., Apex, NC) was used for all the TFF experiments. An OPTISEP 3000 module containing 0.186 m2 of a modified polysulphone membrane with a pore size of 0.8 µm and channel height of 0.5 mm was used for the clarification and protein passage. The protein concentration was performed using an OPTISEP 3000 module containing 0.093 m2 of a regenerated cellulose ultrafiltration membrane with a rating of 100 kD and channel height of 0.75 mm.

At the start of the clarification, a 2-L tank was filled with culture harvest. As permeate left the system, the tank was filled with culture harvest to maintain a constant tank volume until the entire 10 L of culture harvest had been added. The retentate flow rate was set at 1.6 liters per meter (LPM), which corresponds to a shear rate of 1400 sec–1 . After a 10X concentration was reached, the retentate was diafiltered with 2 L of tris-buffered saline (TBS) to remove any small particles or proteins remaining in the medium from the concentrated cells. As the diafiltration continued, the retentate flow rate was slowly ramped up until a shear rate of 7,000 sec–1 was reached.

The SARS-CoV spike protein was extracted from the cells by adding 150 mL of 10X extraction buffer (10% Triton 100 in TBS) to the tank containing the 1.5-L culture harvest after diafiltration. The concentrated cells were circulated in the extraction buffer for 30 min. with the permeate port closed to permit ample mixing and extraction time. The extracted protein was then passed through the same membrane that was used for clarification. The retentate flow rate started at 3.2 LPM and then increased to 7.2 LPM, which corresponds to a shear rate of 6,000 sec–1 . To increase the protein yield, a 1X diafiltration was performed.

The 1.2 L of SARS-CoV spike protein was concentrated using ultrafiltration. The retentate flow rate was set at 12.5 LPM, which corresponds to a shear rate of 6,700 sec–1 . The back pressure was maintained between 20 and 30 psi. To ensure the appropriate buffer for loading the HPLC column, a 1X diafiltration was performed using TBS.

Figure 1. Membrane performance during (a) clarification and diafiltration; (b) protein passage; and (c) protein concentration. The closed circles are the total permeate volume for that step. The open circles are the LMH (liters of permeate per square meter of membrane per hour) rates at that time. A Purosep LT-2Q system with an Optisep 3000 holder (NCSRT, Apex, NC) was used for all TFF experiments. The same module with a modified polysulfone membrane with an 0.8 µm pore size was used for the clarification and diafiltration and passing the protein. A module with a regenerated cellulose membrane with a 100-kD pore size was used for the concentrating the protein. In Figure 1a, the concentration (i) and diafiltration (ii) steps are marked. The arrow in Figure 1b indicates the point at which the retentate flow rate increased from 3.2 to 7.2 LPM. The bar in Figure 1c indicates the time at which diafiltration was performed during the ultrafiltration step.

Results and Discussion

Clarification and Protein Extraction

The membrane performance is displayed in Figure 1a for the initial clarification and diafiltration step. The initial permeate flux was 750 mL/min. As expected in a microfiltration process, the flux slowed as a gel layer built up on the membrane. The cell concentration was increased to 10X in 27 min., which is an average LMH (liters of permeate per square meter of membrane per hour) of 109 L m–2 h–1 . After concentrating, diafiltration was performed on the concentrated cells to remove the medium components and place the cells in an appropriate buffer for additional processing. The permeate flux further slowed as diafiltration was performed on the concentrated cells, yet even at the end of the diafiltration a flux rate of 10 L m–2 h–1 was maintained. The overall average flux rate for the concentration and diafiltration step was 41 L m–2 h–1 .

The SARS-CoV spike protein was extracted from the cells using a Triton 100 extraction buffer, which was simply added to the solution in the filtration skid tank. The desired protein passed through the membrane and was separated from the larger cellular components (figure 1b). The protein recovery was increased by performing a 1X diafiltration, which would result in a 75% theoretical yield of the protein. The protein passage step proceeded with an average flux rate of 10 L m–2 h–1 . The large increase seen in the LMH value (the arrow in Figure 1b) occurred when the retentate flow rate was increased as described in the materials and methods section above. However, the permeate flux quickly decreased, returning to the previous steady state value.

Filtering the cells after the extraction buffer was added was expected to proceed more slowly than the initial clarification for two reasons. First, a gel layer would already have formed on the membrane during the clarification. Second, after the extraction buffer was added, cell lysis would have increased the number of particles in the retentate. Some of these newly formed particles would cause additional membrane fouling.

Ultrafiltration to Concentrate Protein

The ultrafiltration step to concentrate the protein with a buffer exchange diafiltration step proceeded with an average flux rate of 35 L m–2 h–1 (Figure 1c). The permeate flux remained constant during diafiltration and then decreased as the protein was concentrated to the final 4X concentration. In a commercial operation, performing diafiltration and concentration before loading the HPLC column would achieve two objectives. First, the buffer could be adjusted to the optimal buffer to load the column, and second, the volume to load the column could be minimized.

Cell Viability Following TFF

One common problem encountered when harvesting insect cells by both centrifugation and TFF is the lysis and injury of the cells as a result of shear forces. To demonstrate that the cells were minimally damaged under the operating conditions for the clarification, viability measurements were taken using a CEDEX and pictures of the cells were taken using a light microscope (Figure 2). The viability of the culture after harvest was 79%. After a 7X concentration was achieved, the viability of the cells was 69%. These results indicate that 87% of the cells that were viable before clarification remained viable after the clarification. When the cells were examined microscopically, no additional cell debris or broken cells were observed after concentrating to 7X (Figure 2). The dearth of broken cells was expected because this TFF design does not use a retentate screen and the operating conditions were set to minimize the shear forces on the cells.

Figure 2. (a) Microscopic analysis of insect cells producing the SARV-CoV Spike protein. The cell viability was 79%. (b) Insect cells producing the SARV-CoV Spike protein after concentration to 7X using SFTFF. The cell viabilty was 69%. No cell lysis was observed.

Protein Analysis

The passage of the targeted protein for each step (clarification, protein extraction, and concentration) was checked using an SDS-PAGE gel followed by a Western blot on the retentate and supernatant samples (see Figure 3, in which the arrow indicates the desired protein). No protein passage was found before the protein extraction (lane 5) or during the concentration step (lane 8). Therefore, no protein was lost through the membrane during these steps. Good passage of the desired protein was observed after the elution buffer was added (lane 6). When comparing the column flow-through (lane 9) to the fraction eluted from the column (lanes 10 through 13, in order of elution), the eluted fraction contained the desired protein purified. Therefore, the desired protein bound and eluted from the nickel column. Thus, the buffer ex-change performed in the TFF resulted in a buffer that was suitable for loading the column, making any additional processing steps unnecessary.

Figure 3. Analysis of retentate and permeate samples by (a) SDS-PAGE and (b) Western blot analysis. The arrow indicates the position of the desired protein. The lanes contain (1) marker protein; (2) harvested cell sample; (3) retentate at 10X concentration; (4) cell retentate after addition of the elutration buffer; (5) permeate before addition of elution buffer; (6) permeate after addition of extraction buffer; (7) permeate from RC 100 kD module; (9) flow-through from loading the nickel column; (10) elution fraction 1 from the nickel column; (11) elution fraction 2 from the nickel column; (12) elution fraction 3 from the nickel colum; and (13) elution fraction 4 from the nickel column.

Increased Speed and Lower Cost

In this study, it was shown that 10 L of culture broth could be clarified and that 75% of the protein could be extracted in 2.5 h. However, this time could be reduced by decreasing the starting volume-to-membrane-area ratio. For example, increasing the membrane area to 0.3 m2 would decrease the process time to less than 2 h. In this work, the protein yield was not optimized. By increasing the number of system diafiltrations to four during the extraction step, the protein recovery could be increased to more than 99%. Based on a continued flux rate of 5 L m–2 h–1 , the additional time needed to achieve a 99% yield would be 2 h. This time could be decreased by increasing the membrane area to 0.3 m2 such that a 99% yield could be obtained in 3.1 h.

Despite the lower flux rate, using only one module has several advantages over using one module each for clarification and protein extraction. First, buying only one module reduces capital expenses. Also, because the module does not need to be cleaned or changed between the clarification and protein extraction steps, the process time necessary to clean or replace the module is saved. In this case, only 30 min. were needed for the extraction buffer and cells to circulate. In contrast, the time to clean, install, and rinse a new module would be approximately 2 h. In addition, in a commercial process requiring good manufacturing practices, cleaning validation would only need to be performed once instead of being performed twice (once for each of two membranes). Over the lifetime of the process, this reduction in cleaning validation could eliminate the need to validate the module hundreds of times. Another advantage of this setup is that the cell solution is not removed from the tank. This reduces the chance of spillage or contamination.


One advantage of using TFF over centrifugation is that TFF scales up linearly.6 Microfiltration TFF is scaled by maintaining a constant ratio of filtrate volume to membrane surface area. The channel height, membrane material, pore size, and shear rate are kept constant. Based on the results obtained in this study, a 1,000-L batch could be harvested and processed in the same manner as described here in 2.5 h using one hundred times the membrane surface area than the initial trial. If a higher protein recovery yield is desired, the operational time would increase to 4.5 h to recover 99% of the protein. This scale up could easily be achieved by using an Optisep 11000 holder with 20 m2 of membrane. A shorter process time could be achieved by increasing the membrane surface area.


A new protocol for the clarification, extraction, and concentration of protein from insect cells has been presented. This procedure is a modified tangential flow filtration process, "SFTFF," which uses a single module for the clarification and extraction of the desired protein. Using one module for these two steps saves both time and material costs. Finally, diafiltration was performed to remove small proteins and medium components from the protein. Thus, the protein can be placed in a buffer that permits loading onto the chromatography column, which saves processing steps.

Quick Recap

RICHARD CHUBET is a scientist at Protein Sciences Corporation, 1000 Research Parkway, Meriden, CT 06450, 203.686.0800, fax 203.686.0268, rchubet@proteinsciences.comJAMES KACMAR, PH.D, is a process development scientist, and HANK KOPF is chief technology officer, both at NCRST, Inc., 1000 Goodworth Dr., Apex, NC 27539, tel. 919.387.8460, fax 919.387.8540,


1. Li W, Moore MJ, Vasilieva N, Sui J, Wong SK, Berne MA, et al. Angio-tensin-convering enzyme 2 is a functional receptor for the SARS coronavirus. Nature 2003; 426:450–4.

2. Lazarte J, Tosi PF, Nicolau C. Optimization of the production of full length rCD4 in baculovirus-infected Sf9 cells. Biotechnol. Bioeng. 1992; 40: 214–217.

3, Schlaeger EJ, Loetscher H, Gentz R. Fermentation scale-up: production of soluble human TNF receptors. In: Blak JM, Schlager EJ, Bernard AR, editors. Baculovirus and recombinant protein production processes. Basel: Roche Ediitions, 1992, p. 201–208.

4. Trinh L., Shiloach J. Recovery of insect cells using hollow fiber microfiltration. Biotechnol. Bioeng. 1995; 48:401–405.

5. Maiorella B, Dorin G, Carion A, Harano, AD. Crossflow microfiltration of animal cells. Biotechnol. Bioeng. 1991; 37: 121–126.

6. Kempken R, Prei mann,A, Berthold W. Clarification of animal cell cultures on a large scale by continuous centrifugation. J. Ind. Microbiol. 1995; 14: 52–57.

7. Van Reis R, Goodrich EM, Yson CL, Frautschy LN, Dzengeleski S, Lutz H. Linear scale ultrafiltration. Biotechnol. Bioeng. 1997; 55: 737–746.

8. Palacio L, Ho CC, Zydney AL. Application of a pore-blockage-cake-filtration model to protein fouling during microfiltration. Biotechnol. Bioeng. 2002; 79: 260–270.

9. Belfort, G, Davis RH, Zydney AL. The behavior of suspensions and macromolecular solutions in crossflow microfiltration. J. Membrane Sci. 1994; 96: 1–58.

Related Content:

Downstream Processing