OR WAIT 15 SECS
Graduate Student at the chemical and biomolecular engineering department, University of California at Los Angeles
Ashutosh Sharma is a process development engineer at Amgen, Inc.
Anurag S. Rathore is a professor in the Department of Chemical Engineering at the Indian Institute of Technology Delhi and a member of BioPharm International's Editorial Advisory Board, Tel. +91.9650770650, firstname.lastname@example.org.
A case study investigated the root cause of failures in sterile filtration by evaluating the effects and interactions of four operating parameters.
Final filtration with a 0.2-μm filter is considered one of the simplest unit operations. Because it is the last step in the process, however, it is a critical step for successful manufacturing. Filter clogging can occur primarily because of molecular aggregation and can result in deterioration of product quality and longer processing times and thus, loss of plant capacity. This article is the 11th in the "Elements of Biopharmaceutical Production" series. It presents a case study investigating the root cause of failures in sterile filtration by evaluating the effects and interactions of four operating parameters: hold temperature, concentration, hold time, and pH. It is shown that the use of statistical modeling can be effective in solving such problems.
The problem of protein aggregation and filter fouling has been examined extensively using bovine serum albumin (BSA) as a model molecule.1–6 It has been shown that the mechanism behind aggregation in filters is governed by the unique properties of the molecule and its interaction with the membrane.1,3,7,8 Factors that can influence fouling include isoelectric point (IEP), surface chemical groups, and membrane charge of the protein, as well as environmental conditions. For example, a molecule with more free thiol surface groups than the average protein may be more likely to aggregate and foul the filter.3 Saksena and Zydney examined the separation of immunoglobulins and albumin and found that the solution environment has a strong impact on the electrostatic interactions between proteins and the filter.9 Andya, et al., have discussed aggregate formation and stabilization of a humanized monoclonal antibody (MAb) formulation.7
Maa and Hsu have investigated fouling during sterile filtration of recombinant human growth hormone.10,11 They examined several fouling mechanisms including pore constriction, adsorption because of nonspecific binding between protein and membrane, shear-induced adsorption, and hydrophobic interface-induced aggregation and adsorption. The results suggested that hydrophobic interface-induced aggregation and adsorption was the dominant cause of filter fouling. Adding of a surfactant or increasing the pH alleviated the problem, whereas increase in ionic strength resulted in more fouling. Many conditions pertaining to the separation process can have an effect on the filterability of a given solution. These conditions include, but are not limited to, concentration, pH or ionic strength, component interactions, charge effects, membrane pore size, membrane consistency, and temperature.8,12
This article investigates the impact on filter throughput of four operating parameters: hold temperature, concentration, hold time, and pH. Theoretical models and statistical analysis have been used to gain insight into the performance of the filtration step.
The intermediate blocking law for fitting constant flow rate experimental curves is often used to model normal flow filtration.12,13 This is especially applicable to the application presented here involving sterile filtration with pore blocking because of aggregation.12 It assumes that particles can deposit on any part of the membrane surface and that any particle depositing on a pore plugs the pore completely. The decrease in free surface, dS, is proportional to the free surface, S, and the reduction of the free surface of pores is identical to the probability for a pore to get blocked:
Equation 1 can be integrated as follows:
Further, Darcy's law states that
Equations 2 and 3 can be combined as
in which, ΔP is the pressure differential (psig), ΔP0 is the initial pressure differential (psig), V/A is the throughput (L/m2), ε is the porosity, and σ is the clogging coefficient (m-1).
For our case, we performed a nonlinear least squares fit using Microsoft Excel Solver. A porosity of 0.7 was assumed per the membrane specification from the vendor. The clogging coefficient and initial pressure fitting parameters were determined from the equations mentioned above. The maximum throughput for this study was defined as the value when the pressure differential endpoint reached 30 psig. By rearranging Equation 4, this maximum throughput can be determined as follows:
If data on filter performance is available, Equation 5 also can be used to predict the maximum throughput that can be achieved before the pressure drop reaches 30 psig.
A MAb product was used in this study. The protein concentration in the feed material was 150 g/L. The starting material was filtered through a 0.2-μm filter before starting the investigation and also after every concentration operation. The pH of the feed was adjusted as necessary using glacial acetic acid. The protein was concentrated using a ProFlux M12 tangential flow filtration system and Pellicon 2 mini filters (Cat#: P2C010C01) from Millipore Corporation (Billerica, MA). Filtration experiments were performed with Durapore type GV filters (Cat#: GVWP04700) from Millipore Corporation. A Beckman DU-600 spectrometer from Beckman Coulter (Fullerton, CA) was used to measure protein concentration. A pH meter Model 720A from Orion was used to measure the pH of the solution during titration. JMP software version 6 from SAS Institute (Cary, NC) was used to perform the statistical analysis.
In a typical experiment, the protein concentration in the feed was adjusted to the desired value by either concentration or dilution with the formulation buffer. Test solutions were mixed during hold times and processing. All experiments were performed at room temperature. A flow rate of 1,000 LMH was chosen as the model flow rate for the Pmax tests. The material was filtered until either the feed was exhausted or the pressure differential reached 30 psig. The operating parameters tested were hold temperature, protein concentration, hold time, and pH. An operating range of 4–22 °C, 40–100 g/L, and 0–72 hours was chosen for the hold temperature, protein concentration, and hold time, respectively.
Figure 1 illustrates a typical pressure response curve for these filtration experiments. As is evident, the data points fit very well with the curve generated using Equation 4. For the case illustrated here, the pressure drop threshold of 30 psi was reached. In some of the experiments, the pressure drop threshold was not reached. For those cases, Equation 5 was used to calculate the theoretical maximum throughput. Curves for all the experiments yielded an R2 value >0.95 demonstrating the validity of the model used.
JMP analysis was performed on the dependence of initial pressure on the various experimental conditions under consideration. The results shown in Figure 2, indicate that concentration and pH are the two factors that have statistically significant effects (p < 0.05). Concentration also has an effect that is significant in magnitude. This is expected because the concentration of the protein has a direct impact on the initial pore plugging on the membrane surface. Although temperature is expected to affect the initial pressure by its impact on viscosity, the effect seems negligible in Figure 2 because of the noise in the data and a ten-fold larger effect of concentration.
JMP analysis of data showing dependence of theoretical maximum throughput on the various experimental conditions under consideration is presented in Figure 3. The hold time is the only parameter that has a statistically significant effect. The effect is also most significant in its magnitude. This is further confirmed by calculating the clogging coefficient using Equation 4 and plotting it with respect to the hold time. As shown in Figure 4, the data fit the exponential curve from the equation very well (R2 = 0.9957).
The significant effect of the hold time on the clogging coefficient and the throughput occurs because conditions that lead to filter fouling are similar to conditions that foster molecular aggregation. In sterile filtration, aggregation is the most likely cause of clogging. The intermediate law states that each molecule can deposit on any part of the membrane surface and completely plug the pore.12 This is consistent with the protein monolayer theory proposed by Kelly and Zydney.1,3 Molecules prone to aggregation because of shear deformation or free thiol groups form small aggregates. These particles then randomly deposit anywhere on the membrane surface and serve as nucleation sites for growing aggregates. Passing molecules may then become attached to the aggregates at these nucleation sites. As these sites grow in the pore, the pore clogs completely. When a protein solution is held for extended periods of time, aggregates can form and grow. When the filtering process commences, there are now more "seeds" to serve as nucleation sites for pore plugging.
Figure 5 illustrates the relationship between the theoretical maximum throughput and hold time under the chosen experimental conditions. The maximum throughput follows an exponential decay with increasing hold time. For a given product, concentration, pH, and temperature of processing are fixed. To ensure successful execution of the sterile filtration step, the focus should be on ensuring that adequate filter area is available at the time of processing. The required area depends on the change in the clogging coefficient with hold time. This relationship can be stated in the form of Equation 6:
Where, the constant of proportionality, δ is a decay constant (h-1) and (V/A)max,0 is the maximum throughput at a hold time of 0 h.
This relationship can be used to prevent filter clogging and avoid the resulting delays during manufacturing. For every new product, it is recommended that Pmax tests be performed at hold times of zero and a maximum allowable hold time that may be required during manufacturing. If the filterability of the product is not affected by hold time, the data can be used for filter sizing. For cases where hold time has a significant impact on the filterability of the product stream, it is recommended that a third Pmax experiment be performed at an intermediate hold time value. With three data points, the clogging coefficient can be determined using Equation 6 along with the relationship between theoretical maximum throughput and hold time. The exponential fit can be used to generate the filter area that would be required to process a known amount of material for a given hold time.
There may be cases in which other operating conditions such as pH and temperature have a more critical impact than they did in our application. For those cases, a different model would need to be generated and implemented. In either case, the systematic approach presented here can be effective in avoiding filter clogging and the resulting deterioration of product quality and/or longer processing times and, thus, loss of plant capacity. Also, the same filtration curves may be obtained through completely different fouling mechanisms. However, the intermediate blocking law that is presented here is a reasonable approach toward interpreting the fouling effect on pressure drop, as well as quantifying.
Ashutosh Sharma is senior engineer and Anurag S. Rathore, PhD, is a director of process development at Amgen,Inc.,Thousand Oaks,CA, 805.447.4491, email@example.com Anurag Rathore is also a member of BioPharm International's editorial advisory board. Sean Anderson is a graduate student at the chemical and biomolecular engineering department,University of California at Los Angeles, Los Angeles,CA.
1. Kelly ST, Zydney AL. Mechanisms for BSA fouling during microfiltration. J Mem Sci. 1995;107:115–127.
2. Syedain ZH, Bohonak DM, Zydney AL. Protein fouling of virus filtration membranes: effects of membrane orientation and operating conditions. Biotechnol Prog. 2006;22:1163–169.
3. Kelly ST, Zydney AL. Effects of intermolecular thiol-disulfide interchange reactions on BSA fouling during microfiltration. Biotech Bioeng. 1994;44:972–982.
4. Higuchi A, Kyokon M, Murayama S, Yokogi M., Hirasaki T, Manabe S. Effect of aggregated protein sizes on the flux of protein solution through microporous membranes. J Mem Sci. 2004;236:137–144.
5. Ho CC, Zydney AL. A combined pore blockage and cake filtration model for protein fouling during microfiltration. J Colloid Inter Sci. 2000;232:389–399.
6. Bowen WR, Gan Q. Properties of microfiltration membranes: flux loss during constant pressure permeation of bovine serum albumin. Biotech Bioeng. 1991;38:688–696.
7. Andya JD, Hsu CC, Shire SJ. Mechanisms of aggregate formation and carbohydrate excipient stabilization of lyophilized humanized monoclonal antibody formulations. AAPS PharmSci 5. 2003;(2) Article 10. Available from www.pharmsci.org.
8. Marshall AD, Munro PA, Tragardh G. The effect of protein fouling in microfiltration and ultrafiltration on permeate flux, protein retention and selectivity: A literature review. Desalination. 1993;91:65–108.
9. Saksena S, Zydney AL. Effect of solution pH and ionic strength on the separation of albumin from immunoglobulins (IgG) by selective filtration. Biotech Bioeng. 1994;43:960–968.
10. Maa YF, Hsu CC. Membrane fouling in sterile filtration of recombinant human growth hormone. Biotech Bioeng. 1996;50:319–328.
11. Maa YF, Hsu CC. Investigation on fouling mechanisms for recombinant human growth hormone sterile filtration. J Pharm Anal. 1998;87:808–812.
12. Hlavacek M, Bouchet F. Constant flowrate blocking laws and an example of their application to dead-end microfiltration of protein solutions. J Mem Sci. 1993;82:285–295.
13. Hermia J. constant pressure blocking filtration laws—application to power-law non-newtonian fluids. Trans IChemE. 1982;60:183–87.