Efficient Prion Removal from Gonadotropin Solutions by Nanofiltration Membranes
Can a nanofiltration process be leveraged for removal of prions?
ABSTRACT
Prion diseases, or transmissible spongiform encephalopathies (TSEs), are a group of fatal neurodegenerative disorders that includes scrapie and bovine spongiform encephalopathy (BSE) in animals, and Creutzfeldt-Jakob disease (CJD) in humans. The prion agent shows remarkable resistance to common physiochemical inactivation procedures and conventional chemical disinfectants. Methods that reduce TSE infectivity, such as treatment with a strong base or autoclaving, are not always compatible with maintaining the biologic activity of proteins. Therefore, purification processes that separate therapeutic proteins from agents that might cause TSE infectivity are necessary to reduce the possibility of transmission. It is reasonable to assume that available methods to remove viruses and bacteria during the manufacture of gonadotropins are also capable of removing prions. In this study, the authors explored whether a nanofiltration process used during the extraction process of follicle-stimulating hormone from human urine can be effectively leveraged for removal of prions under conditions used for the manufacture of urine-derived gonadotropins.
Prion diseases, or transmissible spongiform encephalopathies (TSEs), are a group of fatal neurodegenerative disorders that includes scrapie and bovine spongiform encephalopathy (BSE) in animals, and Creutzfeldt–Jakob disease (CJD) in humans (1). The agent responsible for such infections is a protein molecule called PrPSc, a conformational variant of the normal host protein, PrPC. The key event in the pathogenesis of TSEs is believed to be the conversion of PrPC into the pathogenic isoform, PrPSc (2). The infectious agent causing TSEs is the prion, which may be identical to PrPSc, a partially protease-resistant isoform of a membrane glycoprotein termed PrPC (3).
Prion diseases of humans are undoubtedly transmissible and are described as genetic (familial), sporadic, or iatrogenic. Iatrogenic transmission of TSEs has occurred through contaminated surgical instruments and through donations obtained from human tissues (e.g., dura mater, pituitary gland, or cornea) (4–6). Evidence shows that prions, the infectious agents causing TSEs, can colonize organs other than the central and peripheral nervous systems, and can be found in extra-cerebral compartments (7, 8).
While growing evidence suggests that infectivity can be transmitted in blood, the presence of PrPSc in urine and associated infectivity remains to be fully confirmed. However, the presence of the noninfective PrPC in some urinary gonadotropin drugs has been recently reported. This finding was not totally unexpected because of the ubiquity and relative abundance of this physiological protein in various regions of the body (9). Indeed, some papers have reported the presence of the infective PrPSc in the urine of animals in models of disease (10).
For these reasons, the debate regarding the use of urinary gonadotropins and their potential risk of transmission of variant CJD is still ongoing (11). Risk assessments need to be conducted in both blood-derived products and urine-derived products, such as gonadotropins, to determine the actual likelihood of transmitting infection and assess actions for risk mitigation.
The prion agent has several unusual characteristics, including a remarkable resistance to physiochemical inactivation procedures such as heat, ionization, ultraviolet light, microwaves, and irradiation, and to conventional chemical disinfectants such as detergents, alcohol, glutaraldehyde, and formalin (12). Methods that reduce TSE infectivity, such as treatment with a strong base or autoclaving, are not in most cases compatible with maintaining the biologic activity of proteins (13). Therefore, purification processes that separate therapeutic proteins from TSE infectivity are necessary to reduce the possibility of TSE transmission by biological components.
Several techniques remove or inactivate PrPSc throughout the manufacturing processes and are applicable to the production of human menopausal gonadotropins and recombinant follicle stimulating hormone (rFSH). Stringent regulatory measuresare already in place to account for viruses and bacteria during the manufacture of gonadotropins, and it is reasonable to assume that such steps can remove significant amounts, if not all, of PrPSc during the manufacturing process. Trials to remove TSE pathogens from biological materials have been made by several investigators using precipitation, chromatography, and filtration (14–22).
The virus filtration step evaluated in this study is used by the Institut Biochimique SA (IBSA) during the process of extracting FSH from human urine. FSH from human urine undergoes a purification process including precipitation with solvents, chromatography, and ultrafiltration. These steps lead to a process that efficiently minimizes the risk of viral contamination in the final product through inactivation, partition, and removal of the potential pathogenic agents.
This study evaluates the prion removal or inactivation achieved by nanofitration membranes (Viresolve normal flow parvovirus [NFP], EMD Millipore) when used according to IBSA's process conditions for the purification of human-derived gonadotropins.
MATERIAL AND METHODS
The risk of prion contamination is a possibility in all biological products of human origin. Evaluation of a step process can be achieved by spiking a significant amount of prions to the material to be nanofiltered, scaling down the original manufacturing step, and determining prion removal.
Description of virus removal filters
Viresolve NFP filters are designed for the removal of small viruses (i.e., 18 nm diameter and largeer) from highly purified proteins. These protein products are produced from recombinant culture, transgenic animals, tissue or body-fluid extractions, and Cohn plasma fractions. The primary mechanism of removal is size exclusion using a composite polyvinylidene fluoride (PVDF) membrane. The structure of the membrane allows proteins as large as 160 kDa to pass, while retaining small particles and viruses such as parvovirus. A small virus model of φX174, a 28 nm nonenveloped bacteriophage, has been validated for the release of Viresolve NFP membrane and filters.
The Viresolve NFP filter has been demonstrated to clear parvovirus in excess of 4 logs in the presence of various protein solutions. The membranes are used in normal flow filtration (NFF) mode. Filtration can be performed under constant flow, by using a peristaltic pumping system, or under constant pressure.
The operation is similar to that based on the 0.22 μm filters widely used in many laboratories. These filters provide robust clearance of viruses that is relatively independent of operating pressure and protein concentration. The membrane has a composite structure wherein a thin ultrafiltration (UF) layer is cast on top of a microporous substrate. The thin UF skin retains viruses effectively without excessively limiting fluid permeability. Various filter formats are available for optimization trials or for pilot- and industrial-scale production, all of which incorporate the same membrane type.
Filtration methods
For the small-scale filterability experiments, the constant pressure V-max model was used (23). V-max is defined as the maximum product solution volume (L) that can be filtered by 1 m2 of membrane before complete plugging. It is a method for predicting the throughput of filters (capacity = L/m2) based on the gradual pore-plugging model. Gradual pore plugging occurs when colloids or suspended matter collects on the sides of filter pores to gradually block them off, until a state of total occlusion is eventually reached. This gradual blocking of the pores occurs in a distinct geometric pattern.
Membrane pore-size distribution in a virus filter device is represented in Figure 1. The largest pores overlap in size with the smallest viruses. The separation challenge calls for carefully characterizing large pores; the smallest (or functional) pores provide robust virus clearance, while large pores lead to virus leakage.
Figure 1: Pore distribution and size of different types of biological contaminants. (ALL FIGURES ARE COURTESY OF THE AUTHORS)
When gradual pore plugging occurs, the smallest pores are the first to be plugged. For example, when more than 75% of a membrane's pores are plugged, the probability of relatively large molecules or viruses passing through the largest pores of the membrane increases. Hence, it is crucial to define the part of the distribution pore area in which the membrane is working. EMD Millipore recommends that the Viresolve NFP be sized using the V75 approach, which is the capacity reached when the flow rate has declined to 25% of the initial flow rate (i.e., 75% of the membrane is plugged).
Selection of the prion strain
For this study, the Rocky Mountain Laboratories (RML6) mouse-adapted scrapie strain was used. Rodent-adapted prion agents have several advantages for this type of study, including a high-titer source of prion infectivity, a high concentration of pathogenic PrPSc, and, in the case of RML, a cell-culture-based assay to titrate prion infectivity (24). A host animal (i.e., tga20 mice) can be used to assay for infectivity with a relatively short incubation period. Mouse prions represent a low biohazard risk because of the lack of scrapie pathogenicity for humans.
Preparation of brain homogenate
Scrapie brain homogenates (20% w/v) were prepared in 0.32 M sucrose from brains of healthy mice (mock homogenate) or from CD1 mice intracerebrally infected with RML6 (Rocky Mountain Laboratories mouse-adapted scrapie strain). Brains were cut into pieces using a scalpel and subsequently homogenized in a ribolyzer using two bursts of 45 s at maximum speed of 6.5. Samples were then stored in 1 mL aliquots at –80°C. The 20% homogenate was diluted 1:1 in phospate buffered saline (PBS) to produce a 10% homogenate.
Design of experiments
During the filtration study, the input and the filtrate were tested for prion infectivity. The total amount of prion infectivity in the filtrate was then compared with the total amount of infectivity in the spiked input. In this study, three independent scaled-down experiments were conducted to evaluate the nanofiltration step used by IBSA during the extraction of FSH from human urine. Operative conditions were selected according to the actual manufacturing process parameters. Table I provides a summary of the three experiments.
Table I: Summary of three experiments. FSH is follicle stimulating hormone, NFP is normal flow parvovirus, PTA is phosphotungstic acid, and RML is Rocky Mountain Labs.
Experiment 1: Crude scrapie brain homogenate spiked into filtration buffer. Crude scrapie brain homogenate was used as a spike at a final concentration of 0.1% and subjected to nanofiltration. One mL of a 10% RML6 scrapie brain homogenate (SBH) was centrifuged at 500 G for 5 min at 4°C. 800 μL of the supernatant was spiked 1:100 into 80 mL of the filtration buffer. This step produced a final concentration of 0.1% or 10–3 dilution of homongenate. One aliquot (5 mL) was removed and frozen at –80°C (input). A 0.1% dilution in filtration buffer of SBH was subjected to nanofiltration using an OptiScale–25 filter at a constant pressure of 2 bar.
Experiment 2: Purified prion preparation spiked into filtration buffer. Because particles significantly reduced the flow rate of the filtration in the previous spike preparation, purification of the prion infectivity solution using phosphotungstic acid (PTA) precipitation was conducted for the second and third experiments. This step minimized the risk of filter fouling while preserving as much infectivity (i.e., high prion titer) as possible. This solution was subjected to nanofiltration using an Optiscale–25 filter at a constant pressure of 2 bar.
Experiment 3: Purified prion preparation spiked into filtration buffer containing FSH (purification intermediate). In a third experiment, the PTA-precipitated prions were spiked into the FSH purification intermediate (0.07 mg/mL FSH) in filtration buffer. This solution was subjected to nanofiltration using an Optiscale–25 filter at a constant pressure of 2 bar.
Preparation of dilution series for FSH interference assay
Five μL of 10% RML6 were serially diluted 1:10 into 45 μL of 10% healthy brain homogenate. Four μL of the resulting dilutions were then diluted 1:1000 into 4 mL of antibiotic-supplemented Opti-MEM (Invitrogen) with 10% fetal calf serum (OFCS). To each of the 4 mL aliquots of OFCS containing the dilutions of the RML6 brain homogenate, 11 μL of FSH solution F200505/Q were added for a concentration of 0.007 mg/mL FSH. This corresponds to the FSH concentration in the 10–1 diluted input or permeate of Experiment 3. The dilution series was then compared with the dilution series containing serial dilutions of 10% RML6, but no FSH.
SDS–PAGE and Western blotting
PrPSc detection was run on the filtrate from nanofilters by Western blot analysis. 12% NuPAGE BisTris (Invitrogen) gels were used for electrophoresis, which was performed at 180 V for 1 h in 1 x MES running buffer. Wet blotting was performed at 110 V, 250–500 mA for 1.5 h in 1 x transfer buffer. Subsequently, the membrane was blocked in blocking buffer (TBS–T containing 5% Topblock, Juro) for at least 1 h at room temperature and then incubated overnight with the PrP-specific monoclonal mouse antibody POM1 (1:3000 dilution) at 2–8°C. The membrane was washed three times for 10 mineach with fresh TBS–T buffer and then incubated for 30 min with the secondary rabbit antimouse IgG1 (Zymed) conjugated with HRP (1:10,000 diluted) at room temperature. Finally, the membrane was again washed three times for at least 10 min each, and the enzymatic reaction initiated by the addition of Enhanced Chemiluminescence (ECL) Western blotting substrate (Pierce).
Tissue culture infectivity assay
For the infectivity-level detection of filtrate samples, the scrapie cell endpoint assay (SCEPA) was used (25–27). Cells growing in 96-well plates are exposed to serial dilutions of input and filtrate samples from experiments 1, 2, and 3, and propagated for several successive cell splits. Dilutions were performed into OFCS so that the final dilution was not less than the equivalent of 10–4 brain homogenate. Dilutions were done so that the total concentration of brain was 10–4 in each inoculum. Samples were kept at 4 °C until cells were infected.
Wells that received only a single infectious unit become extensively infected, as evidenced by the accumulation of PrPSc in many cells in the well. At an appropriate dilution, only a fraction of the wells receive an infectious unit and score positive, and the infectious titer of the starting material can be determined by the Poisson distribution and reference to standard curves.
Cells used for titration of infectivity
A highly prion-susceptible subclone of mouse neuroblastoma cells (N2A) was used to titrate infectivity. This clone was derived from the original N2A American Type Culture Collection (ATCC) cell line by serial isolation of highly prion-susceptible subclones. Cells were stored in liquid nitrogen, and fresh vials were thawed for each titration assay.
Culturing of cells
Highly susceptible N2A cells were cultured in Optimem-I medium (Invitrogen) containing 10% fetal calf serum (Perbio or Invitrogen) and 1000 units/mL penicillin G sodium and 100 μg/mL streptomycin G (OFCS). Cells were split at least every three days or when confluent at 1:3 to 1:5 dilutions. The medium was exchanged every second day. Cells were grown in an atmosphere of 5% CO2 and saturated humidity at 37°C in a cell-culture incubator.
Preparation of cells for SCEPA
Four days before starting the SCEPA assay, a frozen aliquot (about 3 x 106 cells in 0.5 mL freezing medium) was thawed, centrifuged, and suspended in 10 mL OFCS. Cell suspension was seeded in a 10 cm Petri dish and grown to confluence (3–5 days; change medium after 3 days). At confluence, there were about 6 x 106 cells, enough for 3 assay plates.
Infection of cells with prion samples
Twenty thousand susceptible cells in 200 μL OFCS were plated into a 96-well plate. After 16 h at 37°C, the medium was replaced by a 0.3-mL assay sample in OFCS. After 3 days, the confluent monolayer (about 105 cells) was gently suspended. A 100-μL aliquot of each sample was transferred into a well of a 96-well plate containing 200-μL OFCS and grown to confluence. The cells were split 1:3 three more times as above and subsequently 4 times 1:10 (for experiments 1 and 2) and 5 times for experiment 3 and the FSH interference assay. Then the cells were grown to confluence.
After they reached confluence, 25,000 cells from each well were filtered onto the membrane of an ELISPOT plate, treated with PK (0.5 mg/mL for 90 min at 37 °C), and denatured. Individual infected (PrPSc -positive) cells were detected by immunocytochemistry using alkaline phosphatase–conjugated POM1 mouse anti-PrP. Wells clearly showing positive spots were identified using light microscopy and compared to mock controls.
Calculation of prion titer
From the proportion of negative to total wells, the number of infectious tissue culture units (TCI) per mL input was calculated by the Poisson equation P(o)= e–m; where P(o) is the probability that a well is not infected; m is the number of infectious particles/aliquot inoculum to which the cells are exposed. Therefore lnP(o)= ln [empty wells/ total wells] = –m and m = ln [total wells/empty wells]. To calculate the number of TCI units/mL in the undiluted original inoculum, the TCI units/aliquot were multiplied with the dilution factor. TCI units are different from LD50 units: one LD50 unit is the dose that kills half of the animals (or infects half of the wells), whereas the TCI unit is the dose that contains on average one infectious particle. Usually the TCI is similar to the LD50.
RESULTS
In experiment 1, a premature and strong plugging effect was noted (see Figure 2). No prion infectivity was detected in the permeate (see Figure 3). This result corresponded to a ≥ 5.8 log reduction in infectivity compared wiht the input. The result was also in agreement with the Western blot data, which suggested a PrPSc reduction of at least 3.2 logs (see Figure 4).
Figure 2: Observation of a strong and premature plugging effect.
Figure 3: Scrapie cell endpoint assay results. A: Infectivity titer in log infective tissue culture (TCI) units/mL for input sample (red) and permeate (blue); B: reduction in infectivity titer achieved by nanofiltration for the first experiment.
Figure 4: Western blot after ultracentrifugation of serial dilutions of input compared with the permeate after nanofiltration. No PrPSc was detected in the permeate. This corresponds to a ⥠3.2 log fold reduction in PrPSc.
In Experiment 2, using a pretreatment approach of this preparation in filtration buffer, the authors were able to filter a larger volume (i.e., 80 mL, see Table II). Again, no prion infectivity was detected in the permeate by SCEPA, which corresponded to a reduction of at least 5.7 logs in TCI (see Figure 5).
Table II: Summary of experimental results.
In the third experiment, where the PTA-precipitated prions were spiked into the FSH purification intermediate, 80 mL of the spiked solution was nanofiltered. No prion infectivity was detected in the permeate, which corresponded to a reduction of at least 5.0 logs in TCI (see Figure 6).
Figure 6: Scrapie cell endpoint assay results. A: Infectivity titer in log infective tissue culture (TCI) units/mL for input sample (red) and permeate (blue). B: Reduction in infectivity titer achieved by nanofiltration containing follicle stimulating hormone at a concentration of 0.07 mg/mL in the filtration buffer.
Figure 5: Scrapie cell endpoint assay results. A: Infectivity titer in log infective tissue culture (TCI) units/mL for input sample (red) and permeate (blue). B: Reduction in infectivity titer achieved by nanofiltration for the second experiment.
FSH interference assay
To assess whether the presence of FSH would interfere with the test system for prion infectivity (i.e., SCEPA) an interference assay was performed. For this assay, serial dilutions of the RML6 standard inoculum in cell-culture medium were prepared. The inoculum for each dilution contained the same amount of FSH as is present in a 10–1 dilution of the input or permeate of experiment 3 in the cell culture medium. This result was compared with the same RML6 dilution series without FSH in the inoculum.
The prion titers for the SCEPA in the presence of FSH were similar to the ones in the absence of FSH (see Figure 7). This result indicates that the FSH did not have an influence on the performance of the SCEPA.
Figure 7: Scrapie cell endpoint assay results, follicle-stimulating hormone (FSH) interference assay. Infectivity titer in log infective tissue culture (TCI) units/g RML6 brain in the absence (red) or presence of FSH (blue).
DISCUSSION
In this study, the authors investigated the capacity of Viresolve NFP filters to remove the scrapie prion protein, PrPSc, under the actual conditions used for the manufacture of urine-derived gonadotropins. Spiked preparations were designed to present a serious challenge to the filters.
Western blot assays were used to monitor the partitioning of PrPSc during the first nanofiltration trial. These assays are semi-quantitative indicate the relative levels of PrPSc present in different samples. However, the sensitivity of available assays are limited and they provide only an indirect measure of infectivity.
Results demonstrate that the Viresolve NFP membrane reduced prion infectivity in a given sample by more than 5.0 logs, both in buffer and in gonadothropin solution. Viresolve NFP filtration consistently reduced the level of PrPSc to below the limits of detection of the SCEPA infectivity assay, suggesting that this process step is effective for the removal of prions. Retention of prion protein seems to occur at all the membrane loading level, even when the membrane is more than 90% plugged.
To examine the influence of protein on PrPSc removal, buffer alone was tested. In this situation, the removal of PrPSc by Viresolve NFP filters showed no significant differences. Protein (FSH) solution did not interfere in the infectivity reduction calculation, as demonstrated in the FSH interference assay.
Filtration removal mechanisms are mainly related to size exclusion or adsorption. In a typical biopharmaceutical manufacturing processes, the active compound should be efficiently separated from the pathogen agents without affecting its biological activity or modifying its molecular characteristics during the filtration process. Size exclusion is the mechanism of choice for such a purpose. This separation can be efficiently performed if the active compound and the specific pathogen agent are significantly different in size or molecular weight.
CONCLUSION
The physiological PrPC is present in significant amounts in various regions of the body as a monomer with a molecular weight and molecular features similar to those of gonadotropins. Therefore, low traces of this protein can be found in the final preparation in some purification processes (9). Even a nanofiltration step, if present into the purification process, cannot efficiently separate gonadotropins from PrPC. But because infectivity of PrPSc is strictly connected to the formation of high molecular weight aggregates of the same protein the differences in the molecular dimensions between PrPSc and gonadotropins are sufficient to predict a high level of performance of the nanofiltration step using Viresolve NFP filters (6, 28). Indeed, in spite of the fact that no evidence exists about the real prion removal mechanism by nanofilters, the aggregation status and the differences in the molecular weight between gonadotropins and PrPSc could in principle explain the high value of log reduction of infectivity observed in this experimental study and the efficiency of this nanofiltration step. Together, these observations demonstrate that the nanofiltration technology can significantly increase the level of prion safety of human-derived biological products.
ACKNOWLEDGMENTS
The authors thank Dr. Adriano Aguzzi, director of the Institute of Neuropathology at the University of Zurich, and his team, Dr. Harald Seeger and Audrey Marcel, also from the Institute of Neuropathology at the University of Zurich for scientific and technical collaboration in the study.
Paolo Caccia is director of the biomanufacturing department and Luca Angiolini is head of the protein engineering laboratory, both at the Institut Biochimique SA, Switzerland. Renato Lorenz* is a senior biomanufacturing engineer at Merck Millipore, Italy, and Estelle Zelter is in field marketing in the Europe Virus Safety Solutions group at Merck Millipore, France. *To whom correspondence should be addressed, renato.lorenz@merckgroup.com
REFERENCES
1. S.B. Prusiner et al., Cell 93 (3), 337–348 (1998).
2. S.B. Prusiner, Science 216 (4542), 136–144 (1982).
3. H. Büeler et al., Cell 73 (7), 1339–1347 (1993).
4. P. Brown, M.A. Preece, and R.G. Will, Lancet 340 (8810), 24–27 (1992).
5. A.M. Haywood, N. Engl. J. Med. 337 (25), 1821–1828 (1997).
6. A. Aguzzi and M. Polymenidou, Cell 116 (2), 313–327 (2004).
7. A. Aguzzi, Haematologica 85 (1), 3–10 (2000).
8. A. Aguzzi et al., Cell. Mol. Life Sci. 53 (6), 485–495 (1997).
9. A. Van Dorsselaer et al., PLoS One 6 (3), e17815 (2011).
10. H. Seeger et al., Science 310 (5746), 324–326 (2005).
11. R. Matorras and F.J. Rodriguez-Escudero, Hum. Reprod. 18 (5), 1129–1130 (2003).
12. D.M. Taylor, J. Hosp. Infect. 43, S69–S76 (1999).
13. D.M. Taylor, Dev. Biol. Stand. 75, 97–102 (1990).
14. J. Tateishi et al., Biologicals 29, 17–25 (2001).
15. S.R. Petteway et al., Haemophilia 4, 166 (1998).
16. C.F. Gölker et al., Biologicals 24, 103–111 (1996).
17. M. Blum et al., Biopharm Int. 11 (4), 28–34 (1998).
18. J. Tateishi T. Kitamoto, and H. Hiratani, Lancet 2 (8467), 1299–1300 (1985).
19. J. Tateishi et al., Membrane 18, 367–362 (1993).
20. J. Tateishi, et al., "Removal of the Prion Protein Using Validatable Filter," in Animal Cell Technology: Development towards the 21st Century, B.C. Beuvery, W.P. Zeijlemaker, and J.B. Griffiths, Eds. (Kluwer Academic Publishers, The Netherlands, 1995), pp. 637–640.
21. M. Yunoki et al., Biologicals 36 (1), 27–36 (2008).
22. K.J. Gilligan et al., "The Use of Composite Membrane Size Exclusion Filtration to Remove TSE Infectious Particles From a Model Protein Solution," www.millipore.com/techpublications/tech1/ps1020en00, accessed Nov. 2011.
23. F. Badmington et al., Pharm. Tech. 19 (9), 64–76 (1995).
24. P. Klohn et al., Proc. Natl. Acad. Sci. USA 100 (20), 11666–11671 (2003).
25. J.A. Edgeworth et al., Proc. Natl. Acad. Sci. USA 106 (9), 3479–3483 (2009).
26. J. Falsig et al., Nat. Neurosci. 11 (1) 109–117 (2008).
27. S.B. Prusiner, Science 252 (5012), 1515–1522 (1991).
28. V. Smirnovas et al., J. Biol. Chem. 284 (36) 24233–24241 (2009).
