OR WAIT null SECS
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.
Lot-to-lot variations between raw materials can greatly impact process performance.
Creation and qualification of scale-down models are essential for performing several critical activities that support process validation and commercial manufacturing. As shown in Figure 1, these activities include process characterization and production support studies that are performed to evaluate column and membrane lifetimes, demonstrate clearance of host-cell impurities and viruses, and troubleshoot manufacturing issues. While the underlying fundamentals are relatively the same as those when scaling up, some unique considerations should be taken when scaling unit operations down.1-4 The goal when scaling down is to create a small-scale or lab-scale system that mimics the performance of its large-scale (pilot or manufacturing) counterpart, when both the process parameters are varied within their operating ranges and also when a process parameter deviates outside its operating range. Before it can be used for lab studies, the scale-down model needs to be qualified and its equivalence to large-scale examined. Data from an inaccurate scale-down model could result in conclusions that may not be applicable to large-scale, resulting in an unsuccessful process-validation campaign or continued lot failures in a manufacturing campaign.
This article is divided into two segments. The first part focuses on an upstream unit operation — fermentation. The next segment will cover two downstream unit operations — chromatography and filtration. The combined article is the fifth in the "Elements of Biopharmaceutical Production" series.
Fermentation processes often involve several scales of operation, encompassing inoculum development, seed expansion, and production fermentation. The differences in volumes between the steps in a single fermentation process can be 10X to 100X for the pilot scale, and 1,000 to 100,000X for the production scale. This may cause the fermentation processes to be challenging to scale down and the specific process parameters, vessel geometries, and operational control strategies must be evaluated for each step. Some general guidelines to consider in developing a representative scale-down model follow.
Figure 1. Scale-down Models are Best Utilized for Process Characterization and Production Support
Practitioners use the terms "similar reactor" or "similar vessel geometries" to describe optimal conditions for a scale-down strategy. However, similarity in vessel geometry does not necessarily imply identical systems, although this would be the most attractive option. Instead, geometric similarity means that the overall aspect ratios of each vessel (small vs. large) are close enough to not impact performance. More importantly, the impeller and sparger designs and placements within the vessel are nearly identical.
Methods for assessing process performance should be identical between scales. These include sample-dilution schemes and measurement times for calculating culture optical densities, wet and dry cell-weights, media metabolite levels, and protein expression. For example, if spectrophotometers are used to make culture optical-density (OD) measurements, use the same model in the scaled-down process. If the same model is not available, use a spectrophotometer with the same path length and instrument precision (relative to one used in manufacturing). To verify, use several solution standards of known optical density.
At the small scale, minimize reactor sampling volumes and times as much as possible. Since each sample that is taken in the scale-down process represents a larger percentage of the total process volume, the percentage of loss at the small-scale must be minimized to prevent depletion of culture broth beyond acceptable levels. If the sample size cannot be reduced, then adjust the frequency of sampling. For example, optical-density or wet cell-weight measurements performed every two hr at the manufacturing scale could be done every ten hr in the scale-down process, as long as the sampling does not occur at a crucial time such as at the start or finish time of feed delivery.
Maintenance of equivalent oxygen transfer (kLα) and control of dissolved oxygen across process scales is the most important requirement for most fermentation scale-down strategies (see Table 1 for equations). However, the efficiency of oxygen transfer and control in a production-scale fermentor (10,000 to 100,000 L) is often significantly lower than in laboratory-scale (1 to 100 L). The overall oxygen-control strategy — including sparger design, calibration, and placement within the reactor — should be identical to the large-scale process. If the sparger design is different between scales, then agitation, aeration, and oxygen enrichment may need to be adjusted to provide equivalent oxygen transfer in the scaled-down process.
Table 1. Scale-down Equations for a Typical Fermentation Process
For inoculum development and expansion, conserve the vessel geometries, incubation conditions, and working volumes at each step whenever possible. Therefore, if 2-L baffled shake-flasks are used for seed expansion in the full-scale process, use the same volume and flask type in the scale-down process. All process control setpoints and ranges (temperature, shaker speed, stroke length, pH, and dissolved oxygen) should be the same. If the same equipment is not available, then vessels with different volumes but similar geometries may be used. However, the operational control parameters may need to be adjusted to account for different vessel geometries.
Set up and sterilize all process vessels and tanks according to the current manufacturing procedure for the process. Sterilization temperatures, procedures for probe and flowmeter calibration, and post-use cleaning protocols should be the same as the large-scale process. If extended (or multiple) sterilization cycles are necessary for any piece of equipment, assess the impact on process performance and individual medium component stability.
Use GMP-released raw materials that are identical to those used for the full-scale process. Lot-to-lot variations between raw materials, including master or working cell-bank vials, as well as all media components, antifoam additions, and acid and base stock solutions, can greatly impact process performance. If the same lot of raw materials is not available, have the vendor supply a representative lot based on the certificate of analysis. Similarly, prepare buffers and all bulk media solutions according to the GMP-approved manufacturing procedures for the full-scale process. Match current manufacturing procedures for the order of addition of media ingredients, mode of sterilization and addition (filter, heat or steam), mixing times and temperatures, media hold times, and the preparation of mixing tanks.
After assembling a system with all the similar geometries, it is time to tune the controls. All volume-independent operational control-parameter setpoints should be identical to the center point of the operating ranges of the large-scale fermentation process. The volume-independent parameters include:
If oxygen transfer rates are equivalent, then the dissolved-oxygen control setpoint and vessel backpressure should also be held constant.
Except for agitation, use a linear adjustment for all the volume-dependent operational control-parameter setpoints. The scale factor should be equivalent to the ratio of overall process volumes. For example, if the process volumes are 300 L and 15 L, the scale factor (as a divisor) is 20.
The volume-dependent parameters include:
Set agitation to provide either representative oxygen transfer rate (kLa), tip speed (vT), Reynolds Number (NRE), or power-input per unit volume (P/V), according to the equations listed in Table 2. Assuming equivalent vessel geometries and sparger design, the best bet for agitation is to provide a representative oxygen transfer rate (kLa) between scales. If kLa data are not available at the different scales, then set agitation to provide an equivalent power input per unit volume (P/V). This should result in similar mixing profiles across scales, and thus similar oxygen transfer and dispersion. Scale down by constant tip speed or constant Reynolds Number has also been reported.
The general approach to scale-down model qualification is to run all operating parameters at the center of the operating range of the manufacturing process. Compare the specific performance (output) parameters and process control sensitivity at both scales. Establish the acceptance criteria prior to the scale-down runs, which can be based on statistical analysis of historical large-scale data.
Culture growth is a critical performance parameter for qualifying the scale-down model. The assessment measures are culture optical density, percent solids accumulation, and wet and dry cell-weights. Evaluate culture growth both qualitatively, based on a comparison of growth profiles (log and linear plots), and quantitatively, through a direct comparison of specific growth rates. The equipment and procedures used for these measurements should be identical to those used in the current manufacturing process. In a good qualification, the overall specific growth rate and final cell yield will approximate the large-scale process.
Oxygen consumption is another important performance parameter for scale-down qualification. Similar trends in dissolved oxygen profiles, and oxygen and airflow rates represent comparability in oxygen usage by the cultures at each scale. In addition, by using a mass spectrophotometer, specific oxygen uptake rates (OUR) can be calculated and compared across scales through an analysis of off-gas oxygen levels.
A key comparison involves final product titer and quality, at a speific rate of production, which is defined as the amount of product expressed per unit biomass weight (wet or dry). The analytical assays and methodologies used to measure product titer should be identical between scales. Product quality may be difficult to assess, because it requires the development of a representative downstream scale-down model, including any harvest procedures, which are challenging to replicate at the small scale. Data from large-scale runs should be collected and analyzed in order to understand the inherent variability in the analytical assays before using the assays for scale-down model qualification.
Measure the rates of consumption of nutrients (such as glucose) and accumulation of specific metabolites (ammonia, lactate, acetate) at both scales to demonstrate process similarity. Once again, due to the variability in measurements of these components, the equipment and procedures employed for calculating the rates must be identical to those used in the manufacturing process. Since cellular metabolism can be influenced by slight shifts in culture conditions, the exact metabolite levels may vary across scales. However, the specific rates of consumption and accumulation for each component should be similar.
Process-control sensitivity for dissolved oxygen, pH, temperature, agitation, and feed delivery must be verified at the small-scale. For example, feed delivery data must be collected and analyzed to ensure the flow controller accurately delivers the feed at the specified target rate. If equipment limitations cause a change in feed rate, the performance of the scale-down model will not be representative of the large-scale process.
Other process control issues to consider are maintenance of equivalent heat transfer rates during in-process temperature shifts, differences in heat transfer and generation at different scales, and overall PID control loop response times. The process control profiles do not need to be identical to one another, with respect to the fluctuations around a given setpoint, although the trends in the profiles for all the control parameters should be comparable.
Additions of acid and base for pH control should also be measured and compared at each scale. However, it is important to note that the additions of each component may not scale linearly. In some cases, the base that is used may also be a nitrogen source for the culture. Therefore, its usage is related to the growth of the culture, as well as pH control. In addition, slight shifts in cellular metabolism (e.g. lactate production and consumption), which are not easily predicted or controlled, can result in changes in acid and base delivery, in order to maintain culture pH. In any case, variations in acid or base delivery will affect process volumes and media composition. As a result, it is important to evaluate the differences in acid and base addition at each scale for any impact on cell growth or production.
In general, perform at least three small-scale runs in order to demonstrate reproducibility and to determine the inherent variability in the process. Since the scale-down models may also be used to evaluate process deviations, further qualification of the scale-down model — which would include fermentation trials at the edges of the operating parameter ranges — may need to be performed and compared to similar process data acquired in the large-scale process.
Table 2. Comparison of Performance of Fermentation Scale-down Model with Large-scale Fermentation Process
Table 2 summarizes different performance attributes for small and large-scale fermentations including a comparison of product titer, percent solids accumulation, and specific growth rate.
Figure 2. Comparison of Culture Growth Profiles in Small- and Large-scale Fermentations
Figures 2 and 3 illustrate a successful scale-down of a microbial fermentation process based on culture growth and dissolved oxygen profiles, respectively. While the dissolved oxygen profiles in Figure 3 were slightly different, the rates of oxygen consumption (excluding the fluctuation at t=20 hours) were similar.
Figure 3. Comparison of Dissolved Oxygen Profiles in Small- and Large-scale Fermentations
Creation of scale-down models that meet qualification requirements can be challenging, particularly for upstream unit operations. These models can be of great use when performing experimental studies in an efficient and economical fashion. However, there are considerations unique to each unit operation that apply during scale-down. It must be mentioned that while scale-down studies certainly increase the chances of having a successful validation campaign, they only provide guidance. The actual confirmation must be made at large-scale.
Anurag Rathore, is a principal scientist at Amgen Inc., 30W-2-A, One Amgen Center Drive, Thousand Oaks, CA 91320, 805.447.4491, fax 805.499.5008, email@example.comRaj Krishnanwork for Amgen Inc, Thousand Oaks, CA.Stephanie Tozerwork for Amgen Inc, Thousand Oaks, CA.Dave Smileywork for Amgen Inc., Longmont, CO.Steve Rauschwork for Amgen Inc., Longmont, CO.Jim Seely work for Amgen Inc., Longmont, CO.
1. Dahlgren ME, Powell AL, Greasham RL, and George HA. Development of scale-down technique for investigation of recombinant Escherichia coli fermentations: Acid metabolites in shake flasks and stirred bioreactors,
2. Reynolds T, Boychyn M, Sanderson T, Bulmer M, More J, and Hoare M. Scale-down of continuous filtration for rapid bioprocess design: Recovery and dewatering of protein precipitate suspensions, Biotech. Bioeng. 2003; 83:454-464.
3. Varga EG, Titchener-Hooker NJ, and Dunhill P. Prediction of the pilot-scale recovery of a recombinant yeast enzyme using integrated models, Biotechnol. Bioeng. 2001; 74:97-107.
4. Godavarti R, Petrone J, Robinson J, Wright R, and Kelley BD. Scale-down models for purification processes: Approaches and applications, in process validation in manufacturing of biopharmaceuticals, eds. Rathore AS and Sofer G. Marcel Dekker, New York 2005.