AU2014100888A4 - Virus clearance and protein purification methods - Google Patents

Abstract

Protein purification techniques can be used successfully as virus clearance techniques as well as protein purification techniques. Particular methods show high efficiency when used in combination with other methods, or when performed in particular orders. By combining methods in particular ways, the inventors have established methods which can effectively purify proteins and clear proteins of viruses in a highly efficient manner. The methods of the present invention provide compositions in which proteins are totally or substantially cleared of viruses without unduly compromising protein yield. -33-

Description

VIRUS CLEARANCE AND PROTEIN PURIFICATION METHODS TECHNICAL FIELD This invention is in the field of protein purification and virus clearance. BACKGROUND ART 5 Expression of proteins in cells of particular origins has the potential drawback of contamination by endogenous virus in the cells. For example, CHO cell expressed proteins can be contaminated with endogenous retrovirus like particles and adventitious agents. Many techniques are known for purifying proteins. A number of techniques are also known to be suitable for virus clearance. Some purification techniques and/or virus clearance techniques are not 0 suitable when combined in particular orders or with certain other techniques. Purification and viral clearance methods can be very inefficient if performed in such order or combinations. It is therefore an object of this invention to provide methods for viral clearance and protein purification which cause proteins to be efficiently separated from viral contaminants and other unwanted material. 5 DISCLOSURE OF THE INVENTION The inventors have surprisingly found that particular protein purification techniques can be used successfully as virus clearance techniques. Particular methods show high efficiency when used in combination with other methods, or when performed in particular orders. By combining methods in particular ways, the inventors have established methods which can effectively purify proteins and 0 clear proteins of viruses in a highly efficient manner. The methods of the present invention provide compositions in which proteins are totally or substantially cleared of viruses without unduly compromising protein yield. The invention provides a method for virus clearance of a composition comprising a protein of interest and potentially comprising virus particles, comprising steps of: (a) performing cation 25 exchange chromatography on the composition and collecting an eluate; and (b) reducing the pH of the eluate to a pH of 3.0-4.5 to inactivate any virus particles in the eluate to give a composition comprising the protein of interest. The invention also provides a method for purifying a mammalian cell derived protein of interest from a composition comprising the protein of interest and potentially comprising virus particles, 30 comprising at least two virus clearance steps, wherein the at least two virus clearance steps are different virus clearance steps selected from anion exchange chromatography, cation exchange chromatography, pH inactivation at an acidic pH, anion exchange membrane chromatography, nanofiltration and detergent inactivation. The invention also provides a method for purifying a mammalian cell derived protein of interest from 35 a composition comprising the protein of interest and potentially comprising virus particles, -1comprising at least two virus clearance steps. Of the at least two virus clearance steps, at least one virus clearance step may be a virus removal step and at least one virus clearance step may be a virus inactivation step. Alternatively, the at least two virus clearance steps may comprise at least two virus removal steps, or the at least two virus clearance steps may comprise at least two virus inactivation 5 steps. The invention also provides a process for preparing an immunogenic composition, comprising: (a) purifying a mammalian cell derived protein by a method of the invention as described above and; (b) formulating the purified a mammalian cell derived protein into an immunogenic composition. The invention also provides a method for purifying RSV F from a composition comprising RSV F 0 and impurities, comprising steps of: (a) performing anion exchange chromatography on the composition and collecting an eluate containing RSV F; and (b) performing cation exchange chromatography on the eluate to give a composition comprising purified RSV F. The invention also provides a method for purifying RSV F from a composition comprising RSV F and impurities comprising steps of: (a) performing anion exchange chromatography and 5 ultrafiltration, in either order, on the composition to give a first purified composition containing RSV F; and (b) performing a virus clearance step on the first purified composition. The invention also provides a method for purifying RSV F from a composition comprising RSV F and impurities comprising steps of: (a) performing a step of anion exchange chromatography or performing a step of cation exchange chromatography and collecting an eluate; and (b) reducing the 0 pH of the eluate to a pH of 3.0-4.5 to inactivate any virus particles in the eluate to give a composition comprising RSV F. The invention also provides a method for purifying RSV F from a composition comprising RSV F and impurities comprising performing a step of anion exchange membrane chromatography and collecting an eluate to give a composition comprising RSV F. 25 The invention also provides a method for purifying RSV F comprising a step of nanofiltration, wherein the nanofiltration is performed using a sterile cuprammonium regenerated cellulose membrane filter formed from a bundle of straw-shaped hollow fibers whose walls have a three dimensional web structure of pores comprising voids interconnected by capillaries, such that proteins in solution can permeate the fiber membrane and pass through the filter whereas viruses are retained.. 30 The invention also provides a process for preparing an immunogenic composition, comprising: (a) purifying RSV F by a method of the invention, as described above and; (b) formulating the purified RSV F into an immunogenic composition. The invention also provides an immunogenic composition comprising high molecular weight aggregates of RSV F. 35 The invention also provides a composition comprising RSV F, wherein the composition comprises virus at a safe level, or wherein virus is absent from the composition. A safe level of virus may be -2any value accepted by industrial standards or guidelines. 10-6 virus particles per dose of immunogenic composition (or <1 particle per 106 doses), for example, is considered to be a safe level of virus The invention also provides a purified mammalian cell derived protein (e.g. RSV F) of the invention, 5 for use in medicine. Similarly, the invention provides the use of a purified mammalian cell derived protein (e.g. RSV F) of the invention in the manufacture of a medicament for administering to an animal. Similarly, the invention provides a method for treating an animal, comprising a step of administering a purified mammalian cell derived protein (e.g. RSV F) of the invention to the animal. These uses and methods are particularly helpful for immunogenic compositions, which may be used 0 for raising an immune response in an animal of interest e.g. for immunisation. Protein purification steps Protein purification is used to remove impurities from compositions comprising a particular protein of interest. Different purification steps are used to isolate the protein of interest from non-protein components of the composition, as well as from other contaminant proteins. In compositions 5 comprising a protein of interest in which viruses are present, the protein of interest must also be purified from the virus particles to at least a safe level before the isolated protein can be used for medicinal purposes. Protein purification refers to methods for purifying a protein of interest from a composition comprising the protein of interest and impurities. Impurities may include non-protein material 0 including DNA, cell debris, viruses and virus-like particles and proteins other than the protein of interest. The protein of interest may be any protein. Examples of proteins that may be purified using the methods of the invention are RSV F glycoprotein, HIV gpl20 and CMV proteins. Different combinations of purification methods and methods of virus clearance may be used to _5 improve the efficiency of purification. When a purification step or virus clearance step is performed "after" another step, it may be performed immediately after the previous purification or virus clearance step in the method, i.e. no other purification or virus clearance steps are performed between the first step and the second step, other than steps such as dilution or storage which may take place in between the two steps. 30 Alternatively, other step(s) may be performed between the first and second steps. When a purification step or virus clearance step is performed "before" another step, the first step may be performed as the purification or virus clearance step performed immediately before the second step in the method, i.e. no other purification or virus clearance steps are performed between the first step and the second step, other than processes such as dilution or storage which may take place in 35 between the two steps. Alternatively, other step(s) may be performed between the first and second steps. -3- Virus Clearance Virus clearance is method by which a protein of interest in a composition is separated from any virus particles in the composition. Virus clearance methods may also be purification methods, but alternatively may achieve virus clearance without purifying the protein, e.g. by virus inactivation 5 rather than virus removal. Methods of virus clearance include methods that are performed on compositions which are known or suspected to contain virus particles, as well as compositions in which virus particles are only potentially present. Even if a method is performed on a composition in which no virus particles are actually present, the method is still considered to be a method of virus clearance if the method was performed to clear any virus particles that might have been present. 0 Virus particles may be present in a protein composition due to the culturing system used to produce the protein being an environment where virus particles are present. For example, virus particles may be present in mammalian cell culture systems such as CHO cells. It is extremely important to ensure that virus particles are cleared from the protein of interest to a level at which the protein cannot be harmful. This is particularly important where the protein is to be used in medicine, and specifically in 5 immunogenic compositions. The protein of interest may be any protein. Examples of proteins that may be separated from virus particles using the methods of the invention are RSV F glycoprotein, HIV gp120 and CMV proteins.Virus clearance according to the present invention can be achieved by any means such as by performing protein purification steps which are capable of removing virus particles, or using methods 0 which inactivate rather than removing the virus, without damaging the protein of interest. Different methods provide different levels of virus clearance. The level of virus clearance that can be achieved by each virus clearance step affects the number of steps that are required in the final purification and virus clearance method. It is preferred to use the fewest number of virus clearance steps possible which is capable of producing a safe and pure protein of interest. 5 For proteins in which virus clearance is an important consideration, methods are preferably used in which as much virus as possible is cleared from the protein composition without unduly compromising protein yield. Virus clearance may be used to clear any type of virus including both enveloped viruses and non enveloped viruses. Virus clearance may also refer to methods for the clearance of virus-like particles 30 (VLPs). In certain situations it is necessary for multiple virus clearance steps to be used in order to achieve a required level of certainty of virus clearance for an immunogenic composition. The method may include two, three, four, five or six different virus clearance steps. These different virus clearance steps will typically involve two different techniques rather than two runs of the same techniques 35 using different parameters. The different virus clearance steps may be selected from anion exchange chromatography, which for the purpose of the invention is resin anion exchange chromatography, -4cation exchange chromatography, pH inactivation at an acidic pH, anion exchange membrane chromatography, nanofiltration and detergent inactivation. Virus clearance technologies Different virus clearance assays are used depending on the type of virus clearance that the method to 5 be assayed is expected to achieve. As explained above, virus clearance may be achieved by virus removal or virus inactivation. The amount of virus that must be removed by the total protein purification protocol, or the log reduction value (LRV) that must be achieved is determined by the amount of virus present in the composition before purification and viral clearance, and the amount of virus that can be tolerated as a 0 safe level in the final composition comprising the protein of interest. The log reduction value of the virus clearance step is then evaluated by calculating: (Input virus titer per unit volume x total input volume) (Output virus titer per volume x total output volume) Virus spiking study 5 The level of virus clearance achieved by particular method steps can be assayed by spiking model viruses into the composition comprising the protein of interest before the particular method step is performed. Model viruses that may be used to assay virus clearance are Murine Leukemia Virus (MuLV) and Murine Minute Virus (MMV). For virus inactivation steps, such as pH inactivation or detergent inactivation, a sample of a 0 composition comprising the protein of interest is spiked with a model virus. The virus clearance step to be tested is performed on the composition and samples are taken from the composition at particular time points. The log reduction value is calculated for each of the samples taken. For virus removal steps, such as anion exchange chromatography, cation exchange chromatography, anion exchange membrane chromatography and nanofiltration, a sample of a composition comprising 25 the protein of interest is spiked with a model virus and filtered by a prefilter. Due to the different sizes of the MuLV and MMV viruses, different size filters may be used for the pre-filtration of the different viruses used in the virus spiking assays. A 0.22pm filter may be used for MuLV and a 0.1pm filter may be used for MMV. The virus clearance step to be tested in performed on the composition and the log reduction value is calculated after the step has been performed. 30 Purification methods A number of methods which are used in the protein purification and viral clearance methods of the invention are described below: Cation exchange chromatography Methods for protein purification and/or methods for viral clearance may comprise a step of cation 35 exchange chromatography. -5- The cationic exchange chromatography may be carried out using any suitable cationic exchange matrix. Commonly used cation exchange matrices are resins or beads made from derivatised sugars e.g. agarose or cellulose. For instance, one useful type of resin is carboxymethyl (CM) derivatised resins which are weak cation exchangers, such as CM cellulose reins. CM resins may comprise wet 5 crosslinked, polysaccharide beads 45-165pm in diameter which are pre-swollen in 20% ethanol (available as CM-Sepharose FFTM). Other commonly used resins are highly cross-linked agarose resins which comprise particles 75pm in size and are multimodal weak cation exchangers (available as CaptoTM MMC; GE Healthcare). The inventors have found carboxymethyl sepharoseTM resins to be particularly suitable, although other resins may be used. An appropriate amount of resin for the 0 amount of material to be purified can be determined by routine experiments without undue burden. Typical buffers which the protein composition is dissolved in when it is applied to the cation exchange chromatography column include those based on maleic acid, malonic acid, citric acid, lactic acid, formic acid, butaneandioic acid, acetic acid, phosphoric acid, HEPES and BICINE. The cation exchange chromatography column may be equilibrated prior to addition of the protein 5 composition by washing the column with the same buffer that the protein composition is in when it is applied to the column. Cation exchange chromatography may be performed at an acidic pH. For example, cation exchange chromatography may be performed at a pH between 4.0 and 6.5, between 4.5 and 6.0, or between 6.0 and 6.5 e.g. at pH 4.0, 4.5, 5.0, 5.5, 6.0 or 6.5. The pH is usually lower than the pI of the protein of 0 interest. The cation exchange chromatography step can be performed under conditions that allow "flow through" of the protein of interest, wherein impurities bind to the cation exchange matrix while the protein of interest flows straight through the system into the eluate. The use of these conditions simplifies the purification process, as there is no need to use a mobile phase buffer of increasing 5 ionic strength or increasing pH etc. to elute the protein of interest. Alternatively, the cation exchange chromatography step can be performed such that the protein of interest binds to the cation exchange matrix while impurities flow straight through the system. The protein of interest may then be eluted by altering the pH of the eluent. Eluate fractions containing protein may be determined by measuring UV absorption at 280nm. The 30 composition comprising the protein of interest collected in the eluate is highly purified relative to the preparation before the cationic exchange chromatography step. Multiple eluted fractions containing the protein of interest may be combined before further treatment. The cationic exchange chromatography step may be repeated, e.g. 1, 2, 3, 4 or 5 times. For example the composition may be applied multiple times to the same chromatography column, or may be 35 applied to a different chromatography column after passing through the first chromatography column. However, typically the cationic exchange chromatography step will be carried out once. -6- Cation exchange chromatography is useful as a virus clearance step as well as a protein purification step because it is useful for virus removal. Anion exchange chromatography Protein purification methods and/or viral clearance methods may comprise a step of anion exchange 5 resin chromatography. The anionic exchange chromatography may be carried out using any suitable anionic exchange matrix. Commonly used anion exchange matrices are resins or beads made from derivatised sugars (e.g. agarose) such as Q-resins (based on quaternary amines) and DEAE resins (based on diethylaminoethane). The inventors have found that Q-resins which have an average bead diameter 0 of -90pm (available as Q-Sepharose

T

M XL or Q-SepharoseTM FF resins; GE Healthcare) are particularly suitable, although other resins may be used. An appropriate amount of resin for the amount of material to be purified can be determined by routine experiments without undue burden. Typical buffers which the protein composition is dissolved in when it is applied to the anionic exchange chromatography column include N-methyl piperazine, piperazine, L-histidine, bis-Tris, 5 bis-Tris propane, triethanolamine, Tris, N-methyl-diethanolamine, diethanolamine, 1,3 diaminopropane, ethanolamine, piperidine and phosphate buffers. The anion exchange chromatography column may be equilibrated prior to addition of the protein composition by washing the column with the same buffer that the protein composition is in when it is applied to the column. Anion exchange chromatography may be performed at a basic pH. For example, anion exchange 0 chromatography may be performed at a pH between 6.0 and 10.0 or between 7.0 and 9.5 e.g. at pH 6.0, 6.5, 7.0, 7.5, 8.0, 8.5, 9.0, 9.5 or 10.0. The pH is usefully higher than the pI of the protein of interest. The anionic exchange chromatography step can be performed under conditions that allow "flow through" of the protein of interest, wherein impurities bind to the anion exchange matrix while the 5 protein of interest flows straight through the system into the eluate. The use of these conditions simplifies the purification process, as there is no need to use a mobile phase buffer of increasing ionic strength or increasing pH etc. to elute the protein of interest. Alternatively, the anion exchange chromatography step can be performed such that the protein of interest binds to the anion exchange matrix while impurities flow straight through the system. The protein of interest may then be eluted 30 by altering the pH of the eluant. Eluate fractions containing protein may be determined by measuring UV absorption at 280nm. The composition comprising the protein of interest collected in the eluate is highly purified relative to the preparation before the anionic exchange chromatography step. Multiple eluted fractions containing the protein of interest may be combined before further treatment. 35 The anionic exchange chromatography step may be repeated, e.g. 1, 2, 3, 4 or 5 times. For example the composition may be applied multiple times to the same chromatography column, or may be -7applied to a different chromatography column after passing through the first chromatography column. However, typically the anionic exchange chromatography step will be carried out once. Anion exchange chromatography may be performed before cation exchange chromatography ) (described above). Performing these steps in this order provides the advantage that the high pH 5 binding conditions used in anion exchange chromatography can avoid precipitation of particular proteins, e.g. RSV F, at the low pH conditions then used in cation exchange chromatography. This is due to the fact that many host cell proteins have a pI at lower pH range. When pH is close to the pI of the protein, each protein molecule has a low charge or no charge. Therefore protein molecules within a composition do not repel each other, and instead attract each other due to hydrophobic interaction 0 to form complex or aggregation, which may then cause precipitation. The high pH binding conditions used in anion exchange chromatography prevent such precipitation from occurring. Precipitation occurs by complex interactions between host cell proteins and RSVF at lower pH. Therefore, once host cell proteins have been removed using anion exchange chromatography, the composition can be moved to a lower pH without risk of precipitation. 5 Anion exchange chromatography is useful as a virus clearance step as well as a protein purification step because it is useful for virus removal. Ultrafiltration and diafiltration One or more ultrafiltration step(s) may be used to purify the protein of interest by removing low molecular weight species. This step may be used for purification purposes rather than for virus 0 clearance. Ultrafiltration involves the use of hydrostatic pressure to force a liquid against a semipermeable membrane. The filter retains the protein of interest but does not retain solvent or smaller solutes. Continued application of hydrostatic pressure causes the volume of the filtrate to increase, and thus the concentration of the protein of interest in the retentate also increases. Many ultrafiltration 5 membranes are commercially available. The molecular weight cut-off (MWCO) of an ultrafiltration membrane determines which solutes can pass through the membrane (i.e. into the filtrate) and which are retained (i.e. in the retentate). The MWCO of the filter used with the invention will be selected such that most of the protein of interest remains in the retentate. Ultrafiltration may be performed using any filtration method. Possible filtration methods are 30 tangential flow filtration, centrifugal filtration or dead-end filtration. Tangential flow filtration (TFF) involves passing a liquid tangentially across a filter membrane. The sample side is typically held at a positive pressure relative to the filtrate side. As the liquid flows over the filter, components therein can pass through the membrane into the filtrate. Continued flow causes the volume of the filtrate to increase, and thus the concentration of the protein in the retentate 35 increases. TFF contrasts with dead-end filtration, in which sample is passed through a membrane rather than tangentially to it. Many TFF systems are commercially available (e.g. using hollow fibres such as those available from GE Healthcare and Spectrum Labs). The MWCO of a TFF membrane -8determines which solutes can pass through the membrane (i.e. into the filtrate) and which are retained (i.e. in the retentate). The MWCO of a TFF filter used with the invention will be selected such that substantially all of the protein of interest remains in the retentate. The material from which the membrane of the filter is manufactured can impact on the components retained and the rate at 5 which material passes through the membrane. Common membrane chemistries are polysulfone (PS), which is a rigid, high strength polymer and polyethersulfone (PES). PES may be modified to have increased hydrophilicity and to have higher permeate flux rates than un-modified PES. Several different methods are known to transform hydrophobic PES membranes into hydrophilic PES membranes. Often it is achieved by coating the membrane with a hydrophilic polymer. To provide 0 permanent attachment of the hydrophilic polymer to the PES a hydrophilic coating layer is usually subjected either to a cross-linking reaction or to grafting. In methods that do not rely on coating, PES can be dissolved in a solvent, blended with a soluble hydrophilic additive, and then the blended solution is used for casting a hydrophilic membrane e.g. by precipitation or by initiating co-polymerization. Hybrid approaches can be used, in which hydrophilic additives are present during 5 membrane formation and are also added later as a coating. An ideal hydrophilic PES membrane can be obtained by treatment of PES (hydrophobic) with PVP (hydrophilic). Treatment with PEG (hydrophilic) instead of PVP has been found to give a hydrophilized PES membrane that is easily fouled (particularly when using a squalene-containing emulsion) and also disadvantageously releases formaldehyde during autoclaving. 0 Known hydrophilic membranes include Bioassure (from Cuno); EverLUXTM polyethersulfone; STyLUX

T

M polyethersulfone (both from Meissner); Millex GV, Millex HP, Millipak 60, Millipak 200 and Durapore CVGLO1TP3 membranes (from Millipore); FluorodyneTM EX EDF Membrane, SuporTM EAV; SuporTM EBV, SuporTM EKV (all from Pall); SartoporeTM (from Sartorius); Sterlitech's hydrophilic PES membrane; and Wolftechnik's WFPES PES membrane. 5 The concentration techniques of tangential flow filtration, centrifugal filtration and dead-end filtration are not mutually exclusive. One or more methods may be performed as part of a single ultrafiltration/diafiltration step. Whichever technique is chosen, the ultrafiltration step preferably increases the concentration of the protein of interest by at least n-fold relative to the initial concentration, where n is 5, 6, 7, 8, 9, 10, 30 12, 15, 20, 25, 30, 35, 40, 45, 50, 60, 70, 80 or more. The ultrafiltration step preferably results in diafiltration of the composition. In diafiltration, solvent and/or microsolutes (e.g. salts) which are removed during ultrafiltration are replaced by new solvent and microsolutes. In general, removal and replacement occur at the same rate and the volume of the composition is thus kept constant. The overall effect of the process is therefore the replacement of 35 original solvent/microsolutes with new solvent/microsolutes. The methods of the invention may thus include a step of diafiltration. -9- Anion exchange membrane chromatography Protein purification methods and/or virus clearance methods may comprise a step of anion exchange membrane chromatography. Anion exchange membrane chromatography was found by the inventors to be an effective virus clearance step that may be used to clear many different types of viruses by 5 virus removal. Anion exchange membrane chromatography is also advantageous in that it is capable of removing trace impurities such as host cell protein, DNA and endotoxins from the composition. Anion exchange membrane chromatography comprises separating a composition on the basis of its charge, as in resin anion exchange chromatography, but the ionic functional groups are covalently attached to a membrane surface rather than to beads. The membrane may be constructed from a 0 hydrophilic polymeric material, to prevent non-specific protein binding. Membrane chromatography allows for operation at high flow rates and the membranes are disposable. Operation at high flow rates is possible due to the elimination of pore diffusion effects. The functional groups are covalently attached directly to the membrane surface, so they are readily available to biomolecules that pass through the membrane. Components of compositions which are 5 separated using anion exchange membrane chromatography will experience convective mass transfer and film diffusion but not pore diffusion as is experienced with gel beads in resin liquid anion exchange chromatography. Membranes may be cleaned and reused, or may be used in a disposable manner such that a membrane is used for a single chromatography step. Using membranes in a disposable manner 0 eliminates the need for cleaning validation. The anion exchange membrane chromatography step may be carried out using any suitable anion exchange membrane. Anion exchange membranes act as filters with anionic functional groups covalently attached to them. Any anion exchange membrane may be used such as those depicted in Table 1 below. Type A membrane is available as Sartobind Q 15 TM (Sartorius), type B membrane is 5 available as Mustang Q CoinTM (Pall) and type C membrane is available as Chromasorb (Millipore). Anion exchange membranes that are used are typically hydrophilic in order to prevent non-specific protein binding occurring at the membrane. Table 1 Linear Volumetric Pre Bed Pore low fow rate column Membrane Volum Size Membrane Functional rate (mL/min) pressure e (mL) (pm) Mcm/hr) limit (MPa) A 0.41 3-5 Reinforced Quaternary 300 24.0 0.5 Cellulose Ammonium B 0.35 0.8 Modified Quaternary 300 12.5 0.5 PES Ammonium C 0.08 0.65 High MW Primary 75 1.0 .3 -10- Polyethylene Amine The pore size of the membrane may be varied. The pore size of the membrane may be between 0.5pm and 5pm (e.g. 0.50, 0.55, 0.60, 0.65, 0.70, 0.75, 0.80, 0.85, 0.90, 0.95, 1.0, 1.5, 2.0, 2.5, 3.0, 3.5, 4.0, 4.5 or 5.0pm) in diameter. If a pore size which is too small is used then flow rate through 5 the membrane can be limited due to increased back pressures. Typical buffers which the protein composition is dissolved in when it is applied to the anionic exchange membrane chromatography membrane include N-methyl piperazine, piperazine, L histidine, bis-Tris, bis-Tris propane, triethanolamine, Tris, N-methyl-diethanolamine, diethanolamine, 1,3-diaminopropane, ethanolamine, piperidine and phosphate buffers. The anion 0 exchange membrane chromatography step can be performed under conditions that allow "flow through" of the protein of interest, wherein impurities including viruses bind to the anion exchange membrane while the protein of interest flows straight through the system into the eluant. The use of these conditions simplifies the purification process, as there is no need to use a mobile phase buffer of increasing ionic strength or increasing pH etc. to elute the protein of interest. Alternatively, the 5 anion exchange membrane chromatography step can be performed such that the protein of interest binds to the membrane while impurities flow straight through the system. The protein of interest may then be eluted by altering the pH and or conductivity of the eluent. Eluate fractions containing protein may be determined by measuring UV absorption at 280nm. The composition comprising the protein of interest collected in the eluate is highly purified relative to the 0 preparation before the anion exchange membrane chromatography step. Multiple eluted fractions containing the protein of interest may be combined before further treatment. The anion exchange membrane chromatography step may be repeated, e.g. 1, 2, 3, 4 or 5 times. However, typically the anion exchange chromatography step will be carried out once. Anion exchange membrane chromatography may be performed after pH inactivation (see below) has 25 been performed. It is advantageous to use a step which achieves both protein purification and virus clearance as one of the later steps in a purification method e.g. because removing virus that has been inactivated by pH inactivation is safer than removing virus which has not been inactivated. Anion exchange chromatography using the Mustang Q Coin has been shown by the inventors to be the most effective anion exchange membrane for virus clearance. Conductivity of the membrane may affect 30 the ability of the membrane to effectively provide virus clearance and although protein recovery is compromised a little by decreasing conductivity, the log reduction value provided was improved 10 fold for the Mustang Q coin when conductivity was halved from 20mS/cm to lOmS/cm. Nanofiltration Nanofiltration was found by the inventors to be an effective virus clearance step that may be used to 35 clear many different types of viruses by virus removal. Nanofiltration is also advantageous in that it -11is capable of removing trace impurities such as host cell protein, DNA and endotoxins from the composition. Nanofiltration involves the use of hydrostatic pressure to force a liquid against a filter and works by the same principle as ultrafiltration as described above. In contrast to ultrafiltration, the filter allows 5 the protein of interest to flow through but captures other impurities such as viruses which are larger than the protein of interest. Many nanofiltration membranes are commercially available. The pore size of a nanofiltration membrane determines which solutes can pass through the membrane (i.e. into the filtrate) and which are retained (i.e. in the retentate). The pore size of the filter used with the invention will be selected such that substantially all of the protein of interest flows through the 0 membrane. The nanofiltration membrane of the invention may have a pore size of 15-180nm (e.g. 15, 20, 25, 30, 35, 40, 45, 50, 55, 60, 65, 70, 75, 80, 85, 90, 95, 100, 105, 110, 115, 120, 125, 130, 135, 140, 145, 150, 155, 160, 165, 170, 175 or 180nm). The membrane used may be a surfactant free cellulose acetate membrane (available as Virosart Minisart; Sartorius Stedim), a dual layer PES (available as VirosolveTM Pro Micro; EMD Millipore) 5 or a cuprammonium regenerated cellulose (available as PlanovaTM; Asahi Kasei). Presence of virus was shown by the inventors to reduce the flux of the membranes used for nanofiltration. This is due to blockage of the pores in the membranes used in nanofiltration. All filters show some flux decay with increased volume due to blockage. However, the flux decay of a cuprammonium regenerated cellulose filter with a pore size of 20nm was shown to be much less than 0 for other filters. Furthermore, the cuprammonium regenerated cellulose filter is demonstrated by the inventors to provide a greater log reduction value than the other filters and also maintained full recovery of the protein of interest. The cuprammonium regenerated cellulose filter therefore appears to be the most effective nanofiltration filter for the purpose of virus clearance. pH inactivation 5 Treatment of a composition comprising a protein of interest at a particular pH may be used to achieve virus inactivation. pH inactivation may be used to inactivate enveloped viruses by virus inactivation. pH inactivation may only be used as a virus clearance step if the protein of interest is stable at the pHs to which the composition is to be exposed. pH inactivation is affected by a number of 30 parameters including pH, temperature, and incubation time. It is important that the particular parameters that are used are capable of effecting virus inactivation but allow the protein of interest to remain stable in the composition. Acidic pHs may be used in pH inactivation to successfully inactivate viruses that are present in the composition. pHs of between 3.0 and 4.5, (e.g. pH 3.0, 3.1, 3.2, 3.3, 3.4, 3.5, 3.6, 3.7, 3.8, 3.9, 4.0, 35 4.1, 4.2, 4.3, 4.4 or 4.5) may be used. The temperature at which the pH inactivation is performed may be between 18'C and 26'C, (e.g. 18, 19, 20, 21, 22, 23, 24, 25 or 26'C). The composition comprising the protein of interest may be incubated at the appropriate pH and temperature for between 5 and 180 -12minutes (e.g. 5, 10, 15, 20, 25, 30, 35, 40, 45, 50, 55, 60, 65, 70, 75, 80, 85, 90, 95, 100, 105, 110, 115, 120, 125, 130, 135, 140, 145, 150, 155, 160, 165, 170, 175 or 180 minutes). pH inactivation may be performed after cation exchange chromatography. This provides the advantage that the eluate from the cation exchange chromatography step has a low pH. A low pH is 5 also required for the pH inactivation viral clearance step. Therefore additional unnecessary pH titration may be avoided. Deter gent inactivation Protein purification methods and/or virus clearance methods may optionally comprise a step of detergent inactivation. Alternatively, protein purification methods and/or virus clearance methods 0 may not include a step of detergent inactivation. Detergent inactivation may be used to inactivate enveloped viruses by membrane disruption. Mammalian host cell system such as CHO cells are known to harbour endogenous retrovirus particles that persist within the cell and are transmitted from generation to generation. These virus particles are enveloped and are susceptible to detergent inactivation. Therefore detergent inactivation 5 may be particular useful in purification of proteins that have been cultured in mammalian host cell systems. Detergent inactivation may be used as a virus clearance step only if the protein of interest is stable in the detergent to which the composition is to be exposed. Detergent inactivation is affected by a number of parameters including detergent, temperature, and incubation time. It is important that the 0 particular parameters that are used are capable of effecting virus inactivation but allow the protein of interest to remain stable in the composition. These parameters can be readily altered to achieve conditions that are suitable for detergent inactivation. Detergent inactivation may be performed using any suitable detergent. A commonly used detergent for detergent inactivation is t-octylphenoxypolyethoxyethanol, (available as Triton X 100). This 5 detergent is a non-ionic detergent and has been observed to inactivate Sindbis, a model enveloped virus, by at least 4 logs at concentrations ranging from 0.3-2% (w/v) within 60 minutes at 22"C. Detergent may be used at a concentration of between 0.1% and 5% (w/v), e.g.0.1%, 0.15% 0.2%, 0.25%, 0.3%, 0.35%, 0.4%, 0.45%, 0.5%, 0.6%, 0.7%, 0.8%, 0.9% 1.0%, 1.2%, 1.4%, 1.5%, 1.6, 1.8% and 2.0% (w/v). The temperature at which the pH inactivation is performed may be between 30 18'C and 26'C, (e.g. 18, 19, 20, 21, 22, 23, 24, 25 or 26'C). RSV F protein The protein of interest is suitably a RSV F glycoprotein. Human respiratory syncytial virus (RSV) causes respiratory tract infections. It is a major cause of lower respiratory tract infections. A prophylactic medication exists for preterm birth (under 35 weeks 35 gestation) infants and infants with a congenital heart defect (CHD) or bronchopulmonary dysplasia (BPD). Treatment is limited to supportive care, including oxygen therapy. RSV is therefore the most important unmet pediatric vaccine need in developed countries. -13- As explained in WO2011/008974, RSV F is an envelope glycoprotein which is responsible for viral penetration, fusion of viral and cellular membranes and fusion of infected cells with surrounding cells (syncytium formation). RSV F can be expressed in CHO-K1 cells. RSV F may comprise the sequence shown in SEQ ID NO:1. RSV F may have a sequence that is at least 70%, at least 75%, at 5 least 80%, at least 85%, at least 90%, at least 95%, at least 96%, at least 97%, at least 98%, at least 99% or 100% identical to SEQ ID NO: 1. RSV F naturally forms trimers (3F) of RSV F monomers. 3F is a promising vaccine candidate for RSV. Viral clearance of RSV F following expression is CHO cells is very important to mitigate the risk of virus contamination in a RSV vaccine. RSV F purification 0 The inventors have found that particular combinations of purification methods provide effective purification of RSV F as well as providing effective viral clearance of the RSV F protein. pH inactivation and/or anion exchange membrane chromatography and/or nanofiltration may be used as viral clearance steps in a RSV F purification method. The inventors have found that performing these viral clearance steps after the purification steps of anion exchange chromatography and/or 5 cation exchange chromatography can provide an efficient purification method. Anion exchange chromatography and cation exchange chromatography may provide a virus clearance function in addition to protein purification, or they may contribute to the purification process by purifying RSV F from other proteins only. RSV F was found by the inventors to be a stable protein at a wide range of conditions. As a result, 0 the purification of RSV F can be achieved under relatively extreme conditions. For example, purification at acidic pHs, in different detergents and/or at high temperatures is possible. The stability of the RSV F protein has allowed the inventors to provide a purification method for RSV F which comprises steps of anion exchange chromatography and/or cation exchange chromatography and pH inactivation at a pH of between 3.0 and 4.5. Such a process allows efficient 5 general purification of RSV F from other proteins and impurities by virtue of the anion and/or cation exchange chromatography and efficient virus clearance by virtue of the pH inactivation. Preferably both anion exchange chromatography and cation exchange chromatography are used in the purification method. Preferably cation exchange chromatography is performed after anion exchange chromatography, and pH inactivation is performed after cation exchange chromatography. 30 The inventors have found that performing ultrafiltration and diafiltration steps after one or more of the other purification steps can provide an efficient purification method. The purification method may include one, two, three or four ultrafiltration/diafiltration steps. RSV F purification methods may comprise the steps of anion exchange chromatography, cation exchange chromatography, pH inactivation, anion exchange chromatography and nanofiltration. One, 35 two, or three ultrafiltration and/or diafiltration steps may be integrated into a purification method including these purification steps. Cation exchange chromatography may be performed after anion exchange chromatography. pH inactivation may be performed after anion exchange chromatography -14and cation exchange chromatography. Anion exchange membrane chromatography and nanofiltration may be performed after anion exchange chromatography, cation exchange chromatography and pH inactivation. Nanofiltration may be performed after anion exchange membrane chromatography. High molecular weight aggregates 5 High molecular weight RSV F species (HMW F) were found by the inventors to form stable associations in the RSV F compositions purified. These stable associations are formed of at least dimers and trimers of 3F. The inventors have carried out characterisation of HMW F in order to understand the structure of the HMW F species. The characterised HMW F species were then tested for impact on the 0 immunogenicity of RSV F when used as a vaccine. The HMW F identified by the inventors was surprisingly found not to interfere with immunogenicity, efficacy or safety of the RSV F compositions produced. HMW F was found to be present following each step in the process of purification of RSV F. The inventors have therefore surprisingly found that purified RSV F compositions which contain HMW F can be used in an immunogenic 5 compositions e.g. as vaccines. The fact that the aggregates were found to be stable and not to interfere with immunogenicity, efficacy or safety of the RSV F compositions produced, allows for purification methods which do not need to remove RSV F aggregates from the composition. This provides the advantage that despite the presence of HMW F in the composition, a simple and robust purification method may be employed. If HMW F needed to be removed additional steps would be 0 required in the purification method which would require more complex operations and would result in lower recovery and robustness of the product. The invention provides a method for analysing a composition which includes RSV F, comprising a step of measuring the proportion of RSV F in the composition which is HMW F and/or measuring the proportion of RSV F in the composition which is unaggregated 3F. _5 Pharmaceutical compositions Proteins purified by the methods of the invention can be useful components of immunogenic compositions e.g. vaccines. For in vivo administration, such compositions should be pharmaceutically acceptable. Such pharmaceutical compositions usually include components in addition the purified proteins described herein e.g. they typically include one or more pharmaceutical 30 carrier(s) and/or excipient(s) and/or immunological adjuvants. Pharmaceutical compositions are preferably in aqueous form, particularly at the point of administration, but they can also be presented in non-aqueous liquid forms or in dried forms e.g. as lyophilisates, etc. Pharmaceutical compositions may include one or more preservatives, such as thimersal or 35 2-phenoxyethanol. Mercury-free compositions are preferred, and preservative-free vaccines can also be prepared. -15- Pharmaceutical compositions can include a physiological salt, such as a sodium salt e.g. to control tonicity. Sodium chloride (NaCl) is typical, which may be present at between 1 and 20mg/ml e.g. 10±2mg/ml or 9mg/ml. Other salts that may be present include potassium chloride, potassium dihydrogen phosphate, disodium phosphate dehydrate, magnesium chloride, calcium chloride, etc. 5 Pharmaceutical compositions can have an osmolality of between 200mOsm/kg and 400mOsm/kg, e.g. between 240-36OmOsm/kg, or between 290-310mOsm/kg. Pharmaceutical compositions may include a protein in plain water (e.g. w.f.i.) but will usually include one or more buffers. Typical buffers include: a phosphate buffer (except in the fifteenth aspect); a Tris buffer; a borate buffer; a succinate buffer; a histidine buffer; or a citrate buffer. Buffer 0 salts will typically be included in the 5-20mM range. Pharmaceutical compositions typically have a pH between 5.0 and 9.5 e.g. between 6.0 and 8.0. Pharmaceutical compositions are preferably sterile. Pharmaceutical compositions are preferably non-pyrogenic e.g. containing <1EU (endotoxin unit, a standard measure) per dose, and preferably <0.1EU per dose. 5 Pharmaceutical compositions are preferably gluten free. Pharmaceutical compositions are suitable for administration to animal (and, in particular, human) patients, and thus include both human and veterinary uses. Pharmaceutical compositions may be prepared in unit dose form. In some embodiments a unit dose may have a volume of between 0.1-1.0ml e.g. about 0.5ml. 0 The invention also provides a delivery device (e.g. syringe, nebuliser, sprayer, inhaler, dermal patch, etc.) containing a pharmaceutical composition of the invention e.g. containing a unit dose. This device can be used to administer the composition to an animal. The invention also provides a sterile container (e.g. a vial) containing a pharmaceutical composition of the invention e.g. containing a unit dose. 25 The invention also provides a unit dose of a pharmaceutical composition of the invention. The invention also provides a hermetically sealed container containing a pharmaceutical composition of the invention. Suitable containers include e.g. a vial. Embodiments of the Inventions 1. A method for virus clearance of a composition comprising a protein of interest and potentially 30 comprising virus particles, comprising steps of: a) performing cation exchange chromatography on the composition and collecting an eluate; and b) reducing the pH of the eluate to a pH of 3.0-4.5 to inactivate any virus particles in the eluate to give a composition comprising the protein of interest. -16- 2. A method for purifying a mammalian cell derived protein of interest from a composition comprising the protein of interest and potentially comprising virus particles, comprising at least two virus clearance steps, wherein the at least two virus clearance steps are different virus clearance steps selected from anion exchange chromatography, cation exchange chromatography, 5 pH inactivation at an acidic pH, anion exchange membrane chromatography, nanofiltration and detergent inactivation. 3. The method of embodiment 2, wherein the at least two virus clearance steps are different virus clearance steps selected from anion exchange chromatography, cation exchange chromatography, anion exchange membrane chromatography, nanofiltration and detergent inactivation 0 4. A method for purifying a mammalian cell derived protein of interest from a composition comprising the protein of interest and potentially comprising virus particles, comprising at least two virus clearance steps. 5. The method of embodiment 4, wherein at least one virus clearance step is a virus removal step and at least one virus clearance step is a virus inactivation step. 5 6. The method of embodiment 4, wherein at least two virus clearance steps are virus removal steps. 7. The method of embodiment 4, wherein at least two virus clearance steps are virus inactivation steps. 8. A process for preparing an immunogenic composition, comprising: a) purifying a mammalian cell derived protein by a method of any one of embodiments 0 1-7 and; b) formulating the purified a mammalian cell derived protein into an immunogenic composition. 9. The method of any one of embodiments 1-8, wherein the protein is RSV F. 10. The method of any one of embodiments 3-8, wherein the protein is HIV gp120. 25 11. A method for purifying RSV F from a composition comprising RSV F and impurities, comprising steps of: a) performing anion exchange chromatography on the composition and collecting an eluate containing RSV F; and b) performing cation exchange chromatography on the eluate to give a composition 30 comprising purified RSV F. 12. A method for purifying RSV F from a composition comprising RSV F and impurities comprising steps of: a) performing anion exchange chromatography and ultrafiltration, in either order, on the composition to give a first purified composition containing RSV F; and 35 b) performing a virus clearance step on the first purified composition. -17- 13. A method for purifying RSV F from a composition comprising RSV F and impurities comprising steps of: a) performing a step of anion exchange chromatography or performing a step of cation exchange chromatography and collecting an eluate; and 5 b) reducing the pH of the eluate to a pH of 3.0-4.5 to inactivate any virus particles in the eluate to give a composition comprising RSV F. 14. A method for purifying RSV F from a composition comprising RSV F and impurities comprising performing a step of anion exchange membrane chromatography and collecting an eluate to give a composition comprising RSV F. 0 15. A method for purifying RSV F comprising a step of nanofiltration, wherein the nanofiltration is performed using a sterile cuprammonium regenerated cellulose membrane filter formed from a bundle of straw-shaped hollow fibers whose walls have a three-dimensional web structure of pores comprising voids interconnected by capillaries, such that proteins in solution can permeate the fiber membrane and pass through the filter whereas viruses are retained. 5 16. A process for preparing an immunogenic composition, comprising: a) purifying RSV F by a method of any one of embodiments 12-16 and; b) formulating the purified RSV F into an immunogenic composition. 17. An immunogenic composition comprising high molecular weight aggregates of RSV F. 18. A composition comprising RSV F, wherein the composition comprises virus at a safe level, or 0 wherein virus is absent from the composition. 19. A purified mammalian cell derived protein obtainable by a method of any one of embodiments 1 7 and 12-16, for use in medicine. 20. Use of a purified mammalian cell derived protein obtainable by a method of any one of embodiments 1-7 and 12-16 in the manufacture of a medicament for administering to an animal. 25 21. A method for treating an animal, comprising a step of administering a purified mammalian cell derived protein obtainable by a method of any one of embodiments 1-7 and 12-16, to the animal. General The term "comprising" encompasses "including" as well as "consisting" e.g. a composition "comprising" X may consist exclusively of X or may include something additional e.g. X + Y. 30 The term "about" in relation to a numerical value x is optional and means, for example, x+10%. BRIEF DESCRIPTION OF THE DRAWINGS Figure 1 - shows a diagram of pH inactivation in SEC chromatograms for control and pH inactivated samples. Figure 2 - shows an SDS-PAGE gel of RSV F trimer in control and boiled and or/ pH inactivated 35 samples. -18- Figure 3 - shows a graph comparing recovery vs. conductivity using ELISA and anion exchange membrane chromatography. Figure 4 - shows a graph comparing recovery vs. conductivity using BCA and anion exchange membrane chromatography. 5 Figure 5 - shows a graph comparing recovery vs. conductivity using peak area integration anion exchange membrane chromatography. Figure 6 - shows a graph comparing volume vs. time with polynomial fit to provide a model of flux decay of each of the nanofilters tested. Figure 7 - shows a graph comparing flux vs. normalised volume plot. 0 Figure 8 - shows a flow diagram for a detergent inactivation protocol. Figure 9 - shows a graph of an MuLV inactivation time curve following detergent inactivation. Figure 10 - shows a protocol and flow diagram for anion exchange chromatography. Figure 11 - shows a graph of a virus spiking chromatograms for anion exchange chromatography. Figure 12 - shows a protocol and flow diagram for cation exchange chromatography. 5 Figure 13 - shows a flow diagram for a pH inactivation protocol. Figure 14 - shows a graph of an MuLV inactivation time curve following pH inactivation. Figure 15 - shows a table comparing the effect of membrane adsorber, pH and conductivity, and a flow diagram for anion exchange membrane chromatography. Figure 16 - shows a table comparing the effect of nanofilters and loading volume and a nanofiltration 0 protocol. Figure 17 - shows a diagram and table comparing flux decay for each of the nanofilters tested. Figure 18 - shows an SDS-PAGE gel for HMW F fractions. Figure 19 - shows a diagram of the thermal stability of HMW F species by differential scanning calorimetry. 25 Figure 20 - shows an EM image of HMW F species. Figure 21 - shows a diagram comparing the immunogenicity of RSV F - Closed symbols represent two weeks post second vaccination; open symbols represent three weeks post first vaccination. 16 samples were tested, from left to right, samples 1-2 and 15-16 were not adjuvanted, samples 3-6 were adjuvanted with MF59, samples 7-14 were adjuvanted with alum. Where the same antigen dose was 30 tested multiple times, e.g. in samples 3 and 5, 4 and 6, samples were taken from separate batches. Figure 22 - shows a diagram comparing the efficacy of protection provided by F trimer and HMW F based on lung pathology (reduction of peribronchiolitis - Closed symbols represent two weeks post second vaccination; open symbols represent three weeks post first vaccination. 16 samples were -19tested, from left to right, samples 1-2 and 15-16 were not adjuvanted, samples 3-6 were adjuvanted with MF59, samples 7-14 were adjuvanted with alum. Where the same antigen dose was tested multiple times, e.g. in samples 3 and 5, 4 and 6, samples were taken from separate batches. Figure 23 - shows a diagram comparing the safety of F trimer and HMW F based on levels of 5 alveolitis - Closed symbols represent two weeks post second vaccination; open symbols represent three weeks post first vaccination. 16 samples were tested, from left to right, samples 1-2 and 15-16 were not adjuvanted, samples 3-6 were adjuvanted with MF59, samples 7-14 were adjuvanted with alum. Where the same antigen dose was tested multiple times, e.g. in samples 3 and 5, 4 and 6, samples were taken from separate batches. 0 MODES FOR CARRYING OUT THE INVENTION Selection of methods for use in RSV F virus clearance testing Prior to RSV F virus clearance testing, potential virus clearance methods were tested for compatibility with RSV F, in the absence of virus spiking. pH inactivation 5 pH inactivation was tested prior to testing for virus clearance to evaluate the stability of RSV F under the conditions required for pH inactivation. Sample purified by size exclusion chromatography was aliquoted into 1 ml samples. The 1 ml samples were titrated to pH 3.0, 3.5, 4.0, 4.5 and incubated for two hours at room temperature. The incubated samples were immediately buffer exchanged into 100 mM sodium phosphate pH 6.8/150 0 mM NaCl using a 30 kDa Vivaspin 20 centrifugal filter. The buffer exchanged samples were normalized to be 0.45 mg/ml and 200 pl of each sample was analysed using SEC, SDS-PAGE, circular dichroism and electron microscopy. The protein concentrations before buffer exchange were very similar among untreated sample and different pH treated samples, ranging from 1.34 to 1.46 mg/ml 25 Levels of aggregation and degradation were analyzed by size exclusion chromatography- high performance liquid chromatography (SEC-HPLC) and SDS-PAGE. The SEC chromatograms in Figure 1 show that the control samples and pH treated samples had very similar profiles. No samples showed visible aggregate or degradant peaks. The retention volumes were all similar at 6.245 to 6.249 minutes. 30 The SDS-PAGE in Figure 2 shows F trimer at non-reducing condition, F monomer at non-reducing and boiled condition and F1/F2 at reduced and boiled condition. There is no difference in the three conditions between the control and pH treated samples. A circular dichroism spectrum of control and pH treated samples showed chromatograms overlapped with each other very well, indicating the protein structure, especially secondary structure was not 35 changed due to pH inactivation. Furthermore, the RSVF spectrums showed the secondary structure is -20composed of predominantly a-helix, with a small region of -sheet and random coil. This experimental finding corresponds with the known RSVF structure. Anion exchange membrane chromatography Anion exchange membrane chromatography was tested to determine the optimum conductivity and 5 type of membrane to allow for minimal RSV F binding. The optimised method was then used in the virus spiking assays (see below). The flowpath was initially flushed with 50mL water. The flowpath was then primed with 40mL 2mM Tris-HCl buffer. Sample was then adjusted depending on the conductivity and membrane used. Sample load was then added to the flowpath. The load was then eluted and samples analysed. Table 2 shows the different conductivities that were tested for each 0 membrane. Table 2 Sartobind Q15 Mustang Q Chromasorb (mS/cm) Run Crmsr m/m (mS/cm) Coin (mS/cm) 5 1 2 2 10 2 5 5 20 3 10 10 30 4 15 15 40 0 5 20 20 50 An RSV F ELISA was performed to confirm the bioactivity and RSV F protein concentration in the load and filtrate solutions. BCA was used to analyse total protein in each solution. Chromatograms 25 were used to obtain 280nm UV absorbance peaks to produce ratios of flow through to eluate peaks. The results for the different conductivities for each membrane were overlayed on a chromatogram to compare the UV curves and exemplify the change in the ratio of flow through AUC (area under the curve) and eluate AUC for each of the runs. The protein recovery for ELISA, BCA and UV 280nm absorbance peak integration are shown in 30 Table 3 below: -21- Table 3 Elution Peak Percentages ELISA RECOVERY BCA RECOVERY Sartobind Q15 Sartobind Q15 Sartobind Q15 Cond Recovery Cond Recovery Cond Recovery 2 45.6% 2 68.1% 2 38.1% 5 64.0% 5 79.0% 5 71.0% 10 80.2% 10 110.0% 10 59.7% 15 91.1% 15 101.2% 15 82.3% 20 95.6% 20 98.2% 20 95.9% Mustang Q Coin Mustang Q Coin Mustang Q Coin Cond Recovery Cond Recovery Cond Recovery 2 46.0% 2 70.3% 2 35.2% 5 63.2% 5 79.3% 5 49.8% 10 77.5% 10 94.5% 10 66.5% 15 89.1% 15 91.3% 15 83.2% 20 94.7% 20 97.2% 20 83.5% Chromasorb Chromasorb Chromasorb Cond Recovery Cond Recovery Cond Recovery 10 49.2% 10 41.2% 10 14.2% 20 66.4% 20 71.0% 20 23.2% 30 76.4% 30 78.7% 30 57.1% 40 84.0% 40 96.0% 40 48.2% 50 89.1% 50 96.7% 50 52.4% Plots of recovery versus conductivity for each of the analysis methods are shown in Figures 3-5. The chromatograms of each run for each membrane show a difference in flow through areas amongst 5 different loading conductivities. The proportional change in the ratio of flow through peak to eluate peak with variable conductivities is visible in the UV curves of the chromatograms. In analyzing the recovery by peak area integration, which is determined by dividing the flow through peak area (mAU x vol) by the total peak area (flow through+eluate), it was observed that the three highest conductivity values for each membrane yielded >75% RSV F recoveries. In the Sartobind Q15 and 10 Pall Mustang Q Coin chromatograms, the 2 mS/cm conductivity flow through UV curves show significant spikes at the end of load. This could either be due to UV drift or potentially RSV F breakthrough. While an increase in conductivity can cause a significant decrease in RSV F binding, it is likely to also provide an effect on virus binding. The larger surface area of viruses will lead to stronger 15 binding, which is minimally affected by slight increases in conductivity. The Sartobind Q15 and Mustang Q coin membranes exhibited a 91-100% recovery by ELISA using a 10-20 mS/cm load. The Chromasorb gave a 79-97% recovery by ELISA with a 30-50 mS/cm load. -22- Nanofiltration Nanofiltration was tested to ensure that filters with pore sizes of 20nm were adequate to allow RSV F trimer to pass through the pores of the filter. Nanofiltration was performed by pressurizing the reservoir to 30 psi (except for Planova reservoir 5 which is pressurized to 14.3 psi). The valve is opened to begin filtration. Volume measurements were taken every two minutes for first ten minutes and every 5 minutes thereafter. Three filters were tested for their suitability for use in the virus clearance testing (Table 4). Table 4 Filter CM Purified RSV F Volume (mL) Sartorius Stedim Virosart Minisart 50 EMD Millipore Viresolve Pro Micro 31 Asahi Kasei Planova 100 0 An RSV F ELISA was used to confirm the bioactivity and RSV F protein concentration in the load and filtrate solutions. BCA was used to analyze total protein of all solutions. The results of the ELISA and BCA tests are shown in Table 5. Table 5 ELISA BCA Input BCA Filtrate ELISA Input Fia Filter Sample (ug/mL) (ug/mL) (ug/mL) Filrt Virosart Q15 Pool 137.4 116.3 76.32541282 61.29598455 Millipore Q15 Pool 137.4 92.8 76.32541282 44.13108979 (Viresolve) Asahi Kasai Q15 Pool 137.4 107.3 76.32541282 79.2363405 Planova 15 Each filter was tested using 100 L/m 2 (10 mL/cm 2 ) solution containing -0.14 mg/mL total protein and the input solution was -55% pure by ELISA. The filtrate volume measurements were plotted versus time for each filter, in order to model the flux decay (see Figure 6). In the case of the Sartorius Virosart shown in the plot above, a second order polynomial fit was used. This equation was then differentiated, which led to the calculation of the 20 flux at each time point. An example of this differentiation is shown in the equation below. V(t) = -0.0022t 2 +0.6871t+0.477 dV Flux- =- 0.0044t +0.6871 dt After the volume data was fit and an equation for flux was developed, the flux over the entire filtration volume was plotted for each filter. This flux was plotted against normalized volume (L/m 2 ) for each of the nanofilters (see Figure 7). -23- The recoveries were calculated over each filtration for both total protein and RSV F. These recoveries are tabulated in Table 6, along with the flux decay over each filter. Table 6 Filter Overall Flux Decay Total Protein RSV F (% Less than initial value) Recovery Recovery Sartorius Stedim 63.3% 91.9% 87.2% Virosart Minisart EMD Millipore 94.0% 67.5% 57.8% Viresolve Pro Micro Asahi Kasei Planova 41.0% 89.9% 119.6% 5 The Planova filter initially produced a filtrate flux of approximately 65 LMH, which was lower than Virosart filter, yet the filter only experienced a 41.0% flux decay. Despite the lower initial flux, the -20% lower flux decay that the Planova filter experienced compared to the Virosart filter is useful in order to maintain a more constant flux over the filtration of the product solution. With high total protein and RSV F recoveries, both the Planova filter and Virosart filter exhibited 0 good performance. From a flux perspective, the Planova filter experienced approximately a 20% lower flux decay throughout the filtration of CM purified RSV F solution than the Virosart filter. Test of Retrovirus Like Particle in RSV F harvest Before testing the virus clearance capabilities of different purification steps, the virus clearance need of RSV F was determined. Two batches of RSV F harvest at 50L scale were tested using BioReliance 5 to determine the concentration of endogenous retrovirus-like particle (RLP) by qRT-PCR. The concentration of RLP in the RSV F harvests was found to be 2.0x10 3 -4.4x10 3 particles/ml Based on these concentrations, the amount of virus clearance that is required can be calculated (see table 7 below). Table 7 Worse case Better case Unit RetroVLP load in harvest * 4.4* 103 2.0* 103 VLP/m Resultant potential load per dose 1.0*10-6 1.0*10-6 VP/dose RSV F amount per dose 45 25 ug RSV F expression 3 50 ug/ml culture Recovery from culture 15% 50% Purified RSV F 0.45 25 ug/ml culture Culture volume/dose 100 1 ml/dose -24- Process virus reduction 4.4*10" 2.0*10 9 Total LRV needed 12 10 Virus clearance steps needed 3 3 > 4 LRV each Total LRV provided >12 >12 A total log reduction value of around 12 must be achieved in order to provide an immunogenic RSV F composition in which virus is cleared to a safe level. Cytotoxicity and viral interference study 5 Before testing the virus clearance capabilities of the different purification steps, the materials to be used in virus spiking study were tested for cytotoxicity and viral interference. This allows dilution factors to be elucidated that do not interfere with the virus assay. Virus spiking assays for RSV F virus clearance steps Two model viruses: xMuLV and MMV were used in the virus spiking study. The virus spiking 0 volume and ratio at each step were summarized in Table 8 below. Where two viruses were tested for a single purification method, they were spiked separately and the steps were performed at different runs. Table 8 Process step Load volume Virus volume (ml) Approximate Virus species (ml) ratio Detergent treatment (with D) 4D 19 1.0 5% xMuLV concentrations Detergent treatment 19 1.0 5% xMuLV (no D) AIEC 100 5.0 5% XMuLV, MMV CIEC 100 5.0 5% XMuLV, MMV Low pH treatment 19 1.0 5% xMuLV 3 pH Low pH control 19 1.0 5% xMuLV AIEMC Ml, M2, M3 16, 70, 80 0.8, 3.5, 4.0 5% XMuLV, MMV 2 conductivity per M Nanofiltration 100, 50, 31 1.0, 0.5, 0.31 1% MMV NF1, NF2, NF3 -25- Detergent treatment As shown in Figure 8, four concentrations of Triton X-100 were tested for xMuLV inactivation in the load material of UFDF1R (concentrated and diafiltrated harvest). Detergent treatment was performed 5 in a buffer of NaCl and Tris-HCl at pH 8.0.The kinetics of inactivation was studied by sampling at five time points: 0, 15, 30, 60 and 120 minutes. The LRV at each time points was calculated after the virus assay. Detergent treatment was first evaluated for effect of process performance. The results from various assays (ELISA, SEC, CM, EM and SDS-PAGE) showed that 0.25%-2.0% triton treatment did not 0 impact RSVF stability. However, post detergent removal, triton bound on RSVF was observed by assays of BCA, SEC, RPC and EM. The presence of triton also decreased the RSVF recovery on Q column from 60% to 50%. The virus inactivation results from spiking study were summarized in Figure 9 and Table 9. In an MuLV inactivation time curve, 0.25%-1.0% Triton X-100 can achieve > 2.67 LRV even from Time 5 0. 2.0% Triton achieved > 1.97 LRV due to cell toxicity at this Triton concentration. As shown in Table 9, 0.25% to 1.0% Triton even achieved > 4.58 LRV at 120 min by large volume TCID assay, which is more sensitive than small volume assay. Table 9 0 Triton X-100 Concentration LRV (120min) 0% (Control) None 0.25% >4.58 ± 0.28 0.5% >4.58 ± 0.28 1.0% >4.58 ± 0.28 25 2.0% >1.97 ±0.28 Detergent treatment was shown to inactivate MuLV very efficiently. It achieved > 4.58 LRV even from Time 0 using 0.25% TX-100. RSVF is also stable after TX-100 treatment in this study. Anion exchange chromatography 30 As shown in Figure 10, the load material (UFDF1R) was spiked with xMuLV or MMV and then filtered by 0.22 or 0.1 pm filter to become Spiked Load II, which was loaded into the 20 ml Q Sepharose FF column for chromatography. Anion exchange chromatography was performed with the composition in a 2mmM Tris-HCl buffer at pH 8.0. Following the chromatography protocol shown in Figure 10, the output, which is the elution peak, was collected and analyzed by TCID 50 to -26determine the virus concentration. The LRV was then calculated from the spiked load II and output data. As shown in Figure 11, the flow through peak was mainly unbound impurity proteins from UFDFR while the eluate contained mainly RSVF protein. The DNA or spiked virus was still bound on 5 column and can be eluted by regeneration and cleaning steps. The RSVF recovery in the control (without virus spiking) was 77.4%. This step achieved >3.36 ± 0.31 LRV for MuLV clearance and 0.75 LRV for MMV clearance. The virus spiking study demonstrated that anion exchange chromatography has virus clearance capability on MuLV, but not on MMV. This may be due to the fact that MuLV has a lower pI (5.8) 0 than MMV (6.2), or the binding of MuLV at Q column may be stronger than binding of MMV. However, considering the binding pH of 8.0, the pI difference may not have a large effect on the interaction. Alternatively the large size of MuLV (70-100 nm, compared with 18-24 nm of MMV) may cause multisite interaction and enhance its binding on column. Cation exchange chromatography 5 As shown in Figure 12, the load material (UFDF2R, diafiltrated anion exchange chromatography eluate) was spiked with xMuLV or MMV and then filtered by 0.22 or 0.1 pm filter to become Spiked Load II, which was loaded into the 20 ml CM-Sepharose FF column for chromatography. Cation exchange chromatography was performed with the composition in a 20mM NaAcetate buffer at pH 5.0. Following the chromatography protocol shown in Figure 12, the output, which is the elution 0 peak, was collected and analyzed by TCID 50 to determine virus concentration. The LRV was then calculated from the spiked load II and output data. The flow through peak was small percentage of unbound RSVF while the eluate was the RSVF product peak. In MuLV spiking run, most of spiked virus flowed through but small part of them was bound to the column and co-eluted with the product. The RSVF recovery in the control was 88.9%. 5 This step achieved >2.36 LRV for MuLV clearance. pH inactivation As shown in Figure 13, three pHs were tested for xMuLV inactivation of the load material of cation exchange chromatography eluate. The kinetics of inactivation was studied by sampling at 6 time points: 0, 15, 30, 60, 90 and 120 minutes. The LRV at each time points was calculated after the virus 30 assay. pH treatment of RSVF at pH 3.5-4.0 for 2 hours did not cause RSVF aggregation or degradation. ELISA did not show apparent change of antigenicity. CD and EM showed protein structure was not affected by this treatment. The virus spiking study results are summarized in Figure 14 and Table 10. The inactivation kinetics 35 at different pH was demonstrated in MuLV inactivation time curve. pH 3.0-4.5 can achieve > 4.19 LRV even from Time 0. pH 3.75 achieved 0.88 LRV at time 0, 4.21 LRV at 15 min and > 4.19 LRV at 30 min. The inactivation was apparently slower at pH 4.0. The LRV was gradually increased from -27- 0 at Time 0 to 4.02 at 120 minutes. As shown in Table 11, pH 3.50 and 3.75 achieved > 4.97 LRV, but pH 4.0 only achieved 4.17 at 120 min by large volume TCID assay. Table 10 Treatment Condition LRV at 120 min 5 Neutral (Control) None pH 3.50 >4.97 ± 0.38 pH 3.75 >4.97 ± 0.38 pH 4.00 4.17 ± 0.45 0 pH greatly affected virus inactivation kinetics. From pH 3.5 to pH 4.0, the lower the pH, the faster the inactivation that was provided. pH 3.75 achieved > 4 LRV at 15 minutes. This pH may be similar to the pH of the cation exchange chromatography eluate. Therefore, the cation exchange chromatography eluate may be directly held for virus inactivation with no or very minor pH adjustment. 5 Anion exchange membrane chromatography As shown in Figure 15, the load material (UFDF3R, diafiltrated cation exchange chromatography eluate) was spiked with xMuLV or MMV, filtered and then loaded into the membrane modules for virus removal. Three membrane adsorbers, each with two conductivities were tested at the same load volume to MV ratio. The product output, which is the flow through peak in AIEMC, was collected 0 and analyzed by TCID 50 to determine virus concentration. The LRV was then calculated from the spiked load II and output data. Specifically, the 1't and 2 "d 50% flow through were collected and sampled respectively and then pooled together as the total flow through. This was used to study the effect of loading capacity on virus clearance. AIEMC was run at flow through mode for virus capture while RSVF protein flowed through the 25 membrane. It was shown that most RSVF flowed through from membrane but a small percentage was bound on membrane and can be subsequently eluted. The RSVF recovery in the flow through was apparently affected by conductivity. 20 mS/cm had smaller bound percentage and higher flow through than 10 mS/cm. The virus clearance and RSVF recovery results were summarized in Table 11 for different 30 conditions. The effect of membrane adsorber, buffer conductivity and loading capacity on LRV and product recovery was demonstrated. The best virus clearance was achieved using Mustang Q coin at 10 mS/cm, with > 4.28 LRV for MuLV and 5.34 LRV for MMV. The RSVF recovery was similar for different membrane adsorbers with 72-78% at 10 or 15 mS/cm and 97-99% at 20 or 30 mS/cm. -28- Table 11 LRV / Product Recovery Chromasorb Mustang Q coin Sartobind Q15 15 mS/cm 30 mS/cm 10 mS/cm 20 mS/cm 10 mS/cm 20 mS/cm MuLV LRV 1 st 50% FT 2.60 2.63 >4.64 3.72 2.72 2.76 2 nd 50% FT 1.90 1.94 4.12 3.46 2.19 2.23 100% FT 2.06 2.08 >4.28 3.56 2.35 2.40 MMV LRV 1st 50% FT 5.66 4.88 5.77 0.49 2.62 0.52 2 nd 50% FT 2.41 2.52 5.16 0.67 2.44 0.26 100% FT 2.61 2.75 5.34 0.57 2.51 0.35 RSV F Recovery FT 72.4% 97.4% 72.5% 98.9% 78.1% 97.0% Early studies showed that pH had little influence on RSV F recovery while conductivity has a large effect (results not shown) on Sartobind Q. In this study, conductivities were chosen based on 5 preliminary screening. As expected, Chromasorb was not affected by conductivity since this membrane utilises both electrostatic interaction from its primary amine group and hydrogen binding from the base matrix. Very similar LRVs were achieved for MuLV or MMV at 15 and 30 mS/cm. Conductivity affected Mustang Q and Sartobind Q15, especially for MMV clearance due to the charge shielding by salt at higher conductivity (20 mS/cm). MuLV binding was not greatly affected 10 by conductivity, possibly due to the potential multisite interaction of the large virus particle and the presence of less charge shielding by salt. Mustang Q at 10 mS/cm achieved > 4.28 LRV for MuLV and 5.34 LRV for MMV, apparently higher than Chromasorb and Sartobind Q 15 with the same or higher conductivity. This is possible due to higher ligand density of Mustang Q due to its smaller pore size (0.8 pm compared to 3-5 pm for 15 Sartobind). Mustang Q has similar pore size to Chromasorb but the Q functional group is may be more accessible than the primary amine on Chromasorb. The loading capacity affected LRV on all the membrane adsorbers and conditions. Due to the limited ligand density on membrane surface, the higher loading could lead to lower binding of virus and less LRV. -29- Nanofiltration As shown in Figure 16, the load material (UFDF3R, diafiltrated cation exchange chromatography eluate) was spiked with MMV, filtered and then loaded into the membrane modules for virus removal. Three nanofilters were tested at constant pressure and the same load volume/ membrane 5 area ratio. The product output, which is the filtrate, was collected and analyzed by TCID 50 to determine virus concentration. The LRV was then calculated from the spiked load II and output data. Specifically, the 0-25%, 0-50% and 0-100% filtrate were sampled respectively to study the effect of loading capacity on virus clearance. The flux decay of three nanofilters was shown in Figure 17. The flux decay was very different from 0 filter to filter. Viresolve reached 100% decay at 20 L/M 2 , Virosart reached 100% decay at 100L/M 2 , while Planova only reached 35-40% decay at 100L/M 2 . Without virus spiking, the flux was generally higher than with virus spiking, as shown on Viresolve and Virosart. However, Planova did not apparent difference in flux with or without virus. The virus clearance results were shown in Table 12. Planova achieved > 7.13 LRV with full recovery 5 of product, while ViroSart and Viresolve achieved 5-6 LRV with lower product recovery of 58% 87%. Table 12 LRV / Product Recovery Sample Planova VireSart Viresolve 25% filtrate >5.24 5.68 5.41 50% filtrate >5.24 5.18 NA 100% filtrate >5.22 4.79 5.41 100% filtrate (large volume) >7.13 5.23 6.36 RSV F recovery 120% 87% 58% Planova nanofilter demonstrated higher LRV and much lower flux decay than other two nanofilters. 20 Since all of the nanofiltration virus spiking assays were run in dead end filtration mode, the lower level of fouling and higher flux demonstrated by the Planova filter may be due to the different material properties of the Planova filter and the pore size control. Characterisation of high molecular weight aggregates (HMW F) HMW F was purified from the purified RSV F composition by size exclusion chromatography. 25 Following analysis by SDS-PAGE the inventors found that the HMW F species having molecular weights of 370kDa and 256kDa were present in the composition (as shown in Figure 18) with the -30- 256kDa species being present in the highest amounts. The HMW F species were confirmed to be RSV F by protein sequencing. The thermal stability and structure of the HMW F species was tested using differential scanning calorimetry. The melting point of the HMW F species was found to be 1.6'C lower than that of RSV 5 F trimer (as shown in Figure 19), showing that the stability of the HMW F species is lower than that of the RSV F trimer species. Circular dichroism did not show structural differences between the HMW F species and RSV F trimer. Electron microscopy showed dimers and trimers of RSV F trimers to be present, possibly connected 0 via a protein stalk (Figure 20). Immunogenicity, efficacy and safety of HMW F Cotton rats were immunized twice, at 0 days and again at 21 days with purified RSV F trimer and HMW F species. Immunogenicity was evaluated three weeks post l't vaccination and two weeks post 2 "d vaccination. By determining F specific IgG levels and RSV F neutralizing titers, purified HMW F 5 was observed to be as immunogenic as purified RSV F trimer (as shown in Figure 21). HMW F did not negatively impact the immunogenicity of purified RSV F. In order to test efficacy of purified HMW F species cotton rats were immunized twice with purified RSV F trimer or HMW F species. The first immunization took place at 0 days and the second immunization took place at 21 days. The rats were challenged with RSV 4 weeks after the second 0 vaccination. Observation of lung viral plaque titers was made 5 days following challenge and demonstrated that both purified RSV F and HMW F species are protective. A second marker for efficacy is based on lung pathology. Efficacious compositions should decrease peribronchiolitis, a characteristic RSV induced inflammation in the lungs. Again, all vaccinated animals were observed 5 days following challenge to have characteristically decreased levels of peribronchiolitis when 5 vaccinated with RSV F or HMW F; (as shown in Figure 22). Evaluation of HMW RSV F for safety is assessed by considering lung pathology. If safe, an increase in alveolitis (an indicator of vaccine enhanced disease) will not be observed. In this study, alveolitis was assessed 5 days following challenge and an increase in alveolitis using HMW F compared to purified RSV F was not observed. 30 -31-

Claims (5)

1. A method for virus clearance of a composition comprising a protein of interest and potentially comprising virus particles, comprising steps of: a. performing cation exchange chromatography on the composition and collecting an eluate; and b. reducing the pH of the eluate to a pH of 3.0-4.5 to inactivate any virus particles in the eluate to give a composition comprising the protein of interest.

2. A method for purifying a mammalian cell derived protein of interest from a composition comprising the protein of interest and potentially comprising virus particles, comprising at least two virus clearance steps, wherein the at least two virus clearance steps are different virus clearance steps selected from anion exchange chromatography, cation exchange chromatography, pH inactivation at an acidic pH, anion exchange membrane chromatography, nanofiltration and detergent inactivation.

3. A method for purifying RSV F from a composition comprising RSV F and impurities comprising steps of: a. performing a step of anion exchange chromatography or performing a step of cation exchange chromatography and collecting an eluate; and b. reducing the pH of the eluate to a pH of 3.0-4.5 to inactivate any virus particles in the eluate to give a composition comprising RSV F.

4. A method for purifying RSV F comprising a step of nanofiltration, wherein the nanofiltration is performed using a sterile cuprammonium regenerated cellulose membrane filter formed from a bundle of straw-shaped hollow fibers whose walls have a three-dimensional web structure of pores comprising voids interconnected by capillaries, such that proteins in solution can permeate the fiber membrane and pass through the filter whereas viruses are retained.

5. An immunogenic composition comprising high molecular weight aggregates of RSV F. -32-