Process Development and Spiking Studies for Virus Filtration of r-hFSH

May 1, 2013
Nikhil Shaligram

Gautam Daftary

Thomas Preuss

Ansari Usman Ali

Horst Ruppach

John Kaundinya

Anita Samagod

Mahesh Gavasane

Bala Raghunath

Subhasis Banerjee

Subhasis Banerjee, PhD, is a manager of process applications at Millipore India Pvt. Ltd.

BioPharm International, BioPharm International-05-01-2013, Volume 26, Issue 5

This study on a recombinant human follicle stimulating hormone demonstrates the use of virus filters to reduce the risk of contamination.


Demonstration of viral clearance is crucial for the recombinant proteins produced from mammalian cell culture. Size exclusion-based filtration is one of the methods for viral clearance valid for different types of mammalian viruses. In this study, the initial hydraulic performance of the virus filtration has enabled the development of a validation protocol and high-log reduction values (LRVs) for an appropriate panel of mammalian viruses (during spiking studies) with the desired throughput. This is essential to minimize the filter area, the cost of this unit operation, and to achieve a robust process. All of the four panel viruses (i.e., MuLV, PRV, Reo-3, and MVM) were evaluated during spiking studies. A high throughput of 1250 L/m2 was obtained with a high flux because of low protein concentration. This was validated for a predefined throughput of 387 L/m2 with high values of LRVs (>5) achieved resulting in the relatively small area of virus filter (0.04m2) for a 50-L batch volume to be processed in 2 h. This study on a recombinant human follicle stimulating hormone (rhFSH) demonstrates the use of virus filters to reduce the risk of contamination and provide a robust process of virus filtration.

Recombinant proteins and monoclonal antibodies (mAbs) produced using mammalian cell culture-based expression systems are required to demonstrate robust virus clearance during downstream purification to meet regulatory guidelines (1). The guidelines also indicate that the process steps identified to provide and/or claim virus clearance should be validated for their ability to clear virus.

Membrane filtration, which is based on the principle of size exclusion, is often employed as one of the techniques for virus clearance in the downstream processing of cell-culture derived biotherapeutics. To determine the appropriate virus filter (type and size) for a particular feed stream, a hydraulic performance study is generally performed using a scaled down model. Such studies are typically performed with the relevant drug-feed material to determine the expected flow rate and throughput (2). After determination of the hydraulic performance, a virus validation study is carried out using a scaled down model. The virus validation study entails spiking an appropriate panel of mammalian viruses (typically four different viruses of different sizes, structure, and type) into the feed material and determining the virus clearance ability (expressed as log reduction value) of the filter for each virus. The validated throughput (litres/square meter), under the spiked condition, also stipulates the maximum design throughput for the virus filter in a manufacturing process. The work embodied in this paper represents the collaborative effort between Bharat Serums and Vaccines Ltd., India (R&D group) and Merck Millipore, India (Biomanufacturing Sciences Network Group) to develop and optimize the hydraulic performance of a virus filtration step. After initial development, efforts were carried out to establish the optimum flux and process throughput, and a good laboratory practice (GLP) virus validation (spiking) study was carried out at Charles River Biopharmaceutical Services, GmbH, Cologne, Germany.


Equipment setup for hydraulic performance

The experimental setup is shown in Figure 1. The apparatus consisted of the pressure vessel to hold the protein solution, along with necessary valves, the Viresolve Virus filter (Vpro, Merck Millipore) device (3.1 cm2, parvovirus reduction filter), digital electronic balance, and a collection vessel for the filtrate. Air pressure regulators were used to control filtration pressure during the tests.

Figure 1: Experimental setup for hydraulic performance and spiking study. (ALL FIGURES ARE COURTESY OF THE AUTHORS)

Description of tests for hydraulic performance

Prior to an experimental run, the VPro Virus filters were wetted with water for injection (WFI) and then conditioned by filtering the appropriate buffer for 5-7 m at a pressure drop of 30-50 psi. At the end of the buffer conditioning step, the normalized buffer permeability (NBP) was recorded and compared to the typical Normalized Water Permeability (NWP) values for the devices before proceeding with the filtration run.

The protein solution, used for the study, was a recombinant human Follicle Stimulating Hormone (r-hFSH, protein concentration 0.1 gm/L) solution from Bharat Serums & Vaccines Ltd. All the buffers and protein solutions were 0.2 μm filtered. After wetting of Vpro filter, the protein feedstock was placed into the pressure vessel and the operating pressure of 30 psi was set accordingly on the pressure regulator. Once drop-wise flow was observed from the Vpro micro (3.1 filtration area) device, the cumulative volume being filtered (V) as a function of time (t) was recorded. Approximately 200 ml of feed volume was used in each trial. Each trial was carried out for at least 60 m unless gross plugging was observed.

Time/volume (t/V) in the y-axis as a function of time (t) in the x-axis was plotted to draw a curve, the slope of this curve was used to calculate the capacity (Vmax) and the reciprocal of the y-axis intercept was used to calculate initial volumetric flow (Qi) (see Equation 1) of the filter based on Vmax gradual pore plugging model (3, 4).

The process filter area requirement (Amin) (under unspiked condition) can be estimated from Equation 1:

Virus spiking study

The virus spiking study was performed at Charles River Biopharmaceutical Services GmbH, Cologne. The virus stock solutions were ultracentrifuged, the virus pellet suspended in phosphate buffered saline (PBS), and subsequently prefiltered: the Murine Leukemia Virus (MuLV) and Pseudo Rabies Virus (PRV) stocks were 0.45μm filtered, the Reo-3 stock 0.22 μm filtered, and the Minute Mice Virus (MVM) 0.1 μm filtered. After prefiltration, samples from the four virus solutions were withdrawn and analyzed for the virus titer.

The virus spiked test item was prepared as follows: 200 mL of separate four aliquots of the r-hFSH protein (protein concentration 0.1 gm/L) was spiked individually in duplicate with 2 mL of each of ultracentrifuged and prefiltered virus stock solutions (spike ratio 1%) of MVM, Reo-3, MuLV, and PRV. The solution was mixed and samples withdrawn from the spiked pool and analyzed for virus titer (spiked test item). The spiked protein solution was additionally prefiltered: MuLV and PRV using 0.45μm, Reo-3 0.22 μm, and the MVM 0.1 μm filter. A sample was withdrawn and analyzed for virus titers (load, prefiltered). An aliquot of the prefiltered spiked protein solution was kept at room temperature for the entire process time to determine the stability of the virus load during virus retentive filtration (hold, prefiltered). The Viresolve Pro test system was set up (Figure 1) and the NWP and NBP were determined to ensure proper wetting of the Viresolve Pro filters. All the Vpro virus filters that were used for the study were chosen from the Vpro validation kit (VPro virus filters of 'Vpro Validation kit' that is available preintegrity tested by Merck Millipore). The filtration was performed at room temperature at 30 psi test pressure, with nitrogen as the pressure source. The operating conditions and feed used were representative of worst-case manufacturing conditions. The individual virus spiked feed solution was loaded into separate test systems and the filtration process started. Each of the filtrate was collected in separate containers that were continuously being weighed on a balance. The filtration was terminated when the desired volumes of at least 120 mL for each of the MuLV, PRV, and Reo-3 (fraction A) spiked feed streams were obtained. No process interruption/flushing during the spiking studies were done, nor it is practiced in manufacturing. For MVM, the prefixed success criteria for filtration was at least 60 mL for first fraction (filtrate fraction A), at least 40 mL for the second fraction (filtrate fraction B), and at least 20 mL for the third fraction (filtrate fraction C). The total mass of each filtrate was determined. After mixing, a sample was withdrawn from each of the filtrate and analyzed for viral titer by endpoint titration and by Large Volume Plating (Fraction A to C) as follows.

Determination of virus titer by endpoint titration

To determine the virus titer of a sample, serial three-fold dilutions were made with cell-culture medium. 100 μl aliquots of each dilution were added to 8 wells of a 96-well-MTP with cells (in 100 μl cell culture medium per well). The cells were cultivated for a specified incubation period. Then, they were inspected microscopically for virus-induced changes in cell morphology.

Large-Volume Plating (LVP)

The detection limit of a sample depends on its volume incubated with the indicator cells. To improve the detection limit, a large volume of the sample was analyzed (LVP). Briefly, 200 μl of the minimal diluted sample was added to a defined number of wells containing the indicator cells in 100 μl cell culture medium. The cells were cultivated for a specified incubation period. Then, they were inspected microscopically for virus-induced changes in cell morphology.

For LVP determination. If virus-induced changes are observed in some wells of the large volume plating (15-50% of all wells) for a sample, the virus titer is calculated according to the Spearman and Kärber formula. It is assumed that a 1:3 higher concentrated dose compared to the highest dose analyzed leads to an infection of all parallel cultures.

Only two reaction rates (i.e., number of virus positive wells divided through the number of wells tested per dilution) are reported: the reaction rate determined by the LVP and the reaction rate of the virtual 1:3 higher concentrated dose (mode C of calculation).

If virus-induced changes are observed in only a few wells of the LVP (< 15 % of all wells) for a sample, the virus titer is calculated according to the following formula (mode D of calculation):

D: pre-dilution factor of the sample

np: number of virus-positive wells

n: number of all wells tested

Vw: sample volume per well (0.2 mL)

If no virus-induced changes are observed for a sample, the virus titer is determined by the Poisson distribution at the 95% confidence limits derived from "Note for guidance on quality of biotechnological products" (5, 6) (mode E of calculation):

p: 0.05

v: tested sample volume in mL

V: process fraction volume in mL

For endpoint titration. The virus titer (TCID50/mL), which causes a positive result in 50% of the tested cultures (TCID50) of an endpoint titration, is calculated according to the method of Spearman and Kärber (mode A of calculation):

Y0: decadic logarithm of highest dilution factor of the sample, which causes the infection of all parallel cultures (= 8 parallel cultures for mode A and B of calculation)

d: decadic logarithm of dilution step (=log10 3)

Pi: observed reaction rate starting at Y0. Observed reaction rate (number of virus positive wells divided through the number of wells tested) per dilution i starting at Y0

v: decadic logarithm of volume conversion factor (= log10(5) for mode A and B, log10 (3 1/3) for mode C of calculation)

Calculation of the standard error:

se: standard error

Pi: observed reaction rate (number of virus positive wells divided through the number of wells tested, per dilution)

ni: number of determinations

d: decadic logarithm of dilution step (=log10 3)

Calculation of the confidence limits:

c: confidence limits

se: standard error

If the highest dose tested in the endpoint titration does not result in the infection of parallel cultures, the virus titer for this sample is calculated according to the Spearman and Kärber formula. It is assumed that a 1:3 higher concentrated dose, compared to the highest dose analyzed, leads to an infection of all parallel cultures (mode B of calculation).

Determination of the reduction

For virus filtration:

Reduction factor of hold:

R: reduction factor

A0: total virus load of load, pre-filtered

An: total virus load of hold, pre-filtered

MuLV, PRV, Reo-3

Reduction factor of virus filtration:

R: reduction factor

A0: total virus load of load, pre-filtered

An: total virus load of filtrate fraction A


Reduction factor of virus filtration (filtrate fraction A):

R: reduction factor

A0: total virus load of load of fraction A, pre-filtered

An: total virus load of filtrate fraction A

Reduction factor of the virus filtration (sum of filtrate fraction A, B, and C):

R: reduction factor

A0: total virus load of load pre-filtered

An: Sum of 'log10 total virus load' of fraction A, B and C

After each filtration the Vpro filters were tested for integrity by gross leak test.


Hydraulic performance study results

Figure 2 summarizes the results of the initial hydraulic performance study with the unspiked feed.

Figure 2: Vpro hydraulic performance study.

Figure 2A shows the V max plot, from which the Vmax, Qi is obtained and subsequently Amin is calculated from equation 1. The minimum area requirement was calculated to be 0.04 m2 for a 50 L batch to be processed in 2 h (with the unspiked feed and without adding any safety factors)—this is equivalent to a process throughput of 1250 liters/m2 (The max throughput or Vmax for the process is 5922 L/m2).

Figure 2B shows the hydraulic flow decay as a function of throughput—projection from the Vmax model is also shown in the figure and it indicates that the VPro virus filter is only marginally plugged in the application.

Virus spiking study results

The virus titers of the four viruses under test remained stable when incubated with the starting material next to the process. The enveloped MuLV and PRV showed a reduction factor of at least 5.37 log10 and 6.01 log10 respectively. For the non-enveloped virus Reo-3 and MVM the reduction factors were at least 6.43 and 6.32, respectively (see Table I). Very low residual infectivity was found in fraction B and C of MVM run1. Two and four out of 384 analyzed wells, respectively, were positive. No residual infectivity was found in any fraction of MVM run 2.

Table I: Virus removal by Vpro virus filter.


The Vpro hydraulic performance study depicted a very high throughput of 1250 Liters/m2 for the protein solution. The high flux and throughputs that were validated during the spiking study permits the design of an economical virus filtration step with a relatively small virus filter area for a given batch size (in the current situation, an area of 0.04 m2 is required to filter a batch of 50 L in 2h). The virus filtration step will be located in downstream purification after a final chromatography step and prior to the final sterile filtration step. Adsorptive prefiltration (Virus prefilter, Merck Millipore) was also evaluated (data not shown) to determine whether such prefiltration would help to increase or further optimize the throughput of the VPro virus filter. Because the protein concentration was quite low and the hydraulic performance study revealed a relatively high throughput and flux for VPro filter alone, the prefiltration option was not pursued further.

With predefined throughput of 387 liters/m2 (120 mL of spiked feed solution filtered through a 3.1 cm2 Vpro micro device), high and robust LRVs (>5-6) were obtained. The current work may possibly be among the earliest of studies that demonstrate the feasibility and robustness of using virus filters to reduce the risk of virus contamination in a recombinant human follicle stimulating hormone purification process. With the data obtained in the scaledown, studies for scaleup were successfully implemented for scaleup for several manufacturing batches.


Virus filtration in downstream purification of recombinant proteins may potentially represent a high-cost unit operation step, often due to the high cost of the virus filters as well as the low filter flow or throughput obtainable with these filters. It would, therefore, pay to devote sufficient time to develop and optimize the process to minimize the filtration area requirements for the process step. The current study, with recombinant human follicle stimulating hormone, demonstrates the development techniques employed to optimize the implementation of a virus filtration step in the downstream purification process.

Mahesh Gavasane is senior manager, Ansari Usman Ali is an executive, Anita Samagod is general manager, John Kaundinya is president, Gautam Daftary is managing director, all at R&D Group, Bharat Serums & Vaccines, Maharashtra. Thomas Preuss is study director, Horst Ruppach is global manager, viral clearance, both at Charles River Biopharmaceutical Services GmbH, Cologne, Germany. Nikhil Shaligram is process development scientist, Bala Raghunath is WW director, Subhasis Banerjee*, is manager, India and Singapore,, all at Biomanufacturing Sciences Network Group, Merck Millipore, Millipore India.

*To whom all correspondance should be addressed.


1. ICH, Q5a Viral Safety Evaluation of Biotechnology Products Derived From Cell Lines of Human or Animal Origin. International Conference of Harmonization of Technical Requirements for Registration of Pharmaceuticals for Human Use (Geneva, 1999).

2. D.R. Asher et al., Bioproces. Intl. 9 (3), 26-37 (2011).

3. W. Kools, Bioprocessing J. 11 (2), 42-47 (2012).

4. B. Raghunath and J. Royce, Bioprocess. Intl. 4 (7), 56-57 (2006)

5. Note for guidance on virus validation studies: The design, contribution and interpretation of studies validating the inactivation and removal of viruses (CPMP/BWP/268/95, February 1996).

6. Note on guidance on quality of biotechnological products: Viral safety evaluation of biotechnology products derived from cell lines of human or animal origin (CPMP/ICH/295/95, October 1997).

Related Content:

Upstream Processing