December 1, 2009 (Vol. 29, No. 21)

Study Demonstrates Correlation between Microbioreactor and Bench-Scale Systems

Current PAT and QbD initiatives are driving the need for a more comprehensive understanding of cell culture processes. The objective of these initiatives is to ensure product quality and performance through the design of effective and efficient manufacturing processes. This can be achieved with a detailed understanding of how various process factors and their interactions affect the product.

The challenge for bioprocess development lies in the limited experimental capacity provided by traditional scale-down models. Simpler models such as shake flasks and well plates do not provide the measurements and controls required to accurately predict the full-scale manufacturing process. As a result, these platforms offer limited applicability for exploring the knowledge space and identifying critical process parameters.

Bench-scale bioreactors are the tried and true scale-down model but have economic and throughput limitations. Characterization of the design space is often left to the final stages of process development, after clone and media selection have already occurred. Therefore, the best clone at manufacturing scale may have been prematurely eliminated, and critical interactions between factors may still be unknown due to limited experimental designs.

In order to fully recognize the benefits of PAT and QbD, a higher throughput scale-down model is required. Such a scale-down model should have measurement and control capabilities similar to bench-scale and larger bioreactors in order to accurately predict performance in the manufacturing process. Seahorse Bioscience has developed the SimCell platform, which is capable of performing hundreds of bioreactor-relevant fed-batch experiments with measurement and control of temperature, pH, dissolved oxygen (DO), and glucose.

The core technology of the SimCell platform is the Bioreactor Card. Each card contains six microbioreactors, each with a working volume of approximately 700 microliters. Each microbioreactor contains immobilized sensors for the noninvasive measurement of pH and DO. Total cell density is measured via forward light scattering.

Bioreactor Cards are constructed of gas-permeable membranes for the exchange of culture gasses (e.g., O2 and CO2) with the incubator environment. Cards are manipulated by a robotic system for automatic inoculation, incubation, mixing, sampling, feeding, and process monitoring and control. Comprehensive software tools enable multifactor experimental designs to be implemented across hundreds of microbioreactors with real-time data tracking. On-line and off-line data sets can be combined for each microbioreactor and analyzed to identify significant trends.

Case Study

In this study, a CHO cell line producing a human IgG1 monoclonal antibody was subjected to a four factor, mixed level design for process optimization. Culture pH and DO were controlled at two different levels, while two feeds were added at three different levels. The full factorial design yielded a total of 36 unique experimental conditions, which were run with either four or six replicates to yield a total of 30 Bioreactor Cards (180 microbioreactors). 

Each culture was run in a 13-day fed-batch process with pH, DO, and glucose control. The temperature for all cultures was 36.5°C and the CO2 was fixed at either 5% or 10%, depending on the pH set point. DO control was achieved by changing the incubator oxygen set point in response to the DO measurement of all microbioreactors residing in that specific incubator.

Control of pH was accomplished by measuring the pH of each microbioreactor, comparing the result to the desired set point, and adding a calculated volume of base for the adjustment. Glucose control was accomplished in a similar fashion except that measurement was performed off-line. In addition to glucose, cell viability and product titer were also measured off-line using samples collected from each microbioreactor.

Specific productivity, defined as terminal titer divided by terminal integrated viable cell concentration (IVCC), was chosen as the performance metric for this study. In order to determine the impact of each process factor on productivity, leverage plots were constructed as shown in Figure 1.

For each plot, the slope of the red line indicates the sensitivity of the process to that particular factor. A horizontal line, as observed for feed 2, indicates no effect, whereas a sloped line, as observed for pH, indicates the process is sensitive to changes in this parameter. For this particular cell line, we observed that the process was sensitive to changes in pH and feed 1, with low pH and high feed rate providing the best results.

Figure 1. Leverage plots show the sensitivity of the process to changes in feed 1, feed 2, pH, and DO.

Scale-Up Study

The next phase of the study sought to verify the results obtained using the SimCell platform in bench-scale bioreactors. For this, a subset of 11 conditions was selected from the factorial design and run in 3 L bioreactors with 1 L working volume, one run per condition. The process parameters used in the bioreactor scale-up study were as close as possible to those used in the microbioreactors given the differences between the platforms.

The correlation between the specific productivity obtained in the microbioreactors and bench-scale bioreactors for all 11 conditions are plotted in Figure 2. In general, the specific productivity measured in the microbioreactor is slightly higher than that of the bench-scale bioreactor. This is partially due to differences in cell quantification between the two platforms. However, from the figure, it is clear that there is a strong correlation between the two scales with a calculated R2 value of 0.90.

In addition to specific productivity, a more detailed comparison was made for one of the conditions run in both microbioreactors and bench-scale bioreactors. For this condition, intact IgG, purity, and percent nonfully glycosylated forms were compared using microchip-SDS on a Caliper Life Sciences LC90 system.

The need for comprehensive process understanding as part of QbD and other initiatives requires large numbers of experiments to be executed. Current scale-down bioreactor models typically do not have the throughput to satisfy these experimental demands.

Simplified scale-down models such as shaken flasks and well plates generally do not have the process measurement and control capabilities to provide comparable data sets. Therefore, a bioreactor scale-down model with higher throughput is required to meet this demand. The SimCell platform offers the ability to execute large statistically designed experiments with automated feeding, measurement, and control comparable to bench-scale bioreactors. The high data content allows statistical analyses to be conducted across a variety of metrics, such as IVCC, titer, specific productivity, and product quality, to identify which factors significantly affect the process.

In the example study presented here, four process factors were varied across a total of 36 unique experimental conditions. Specific productivity was chosen as the primary metric and the SimCell platform identified feed 1 and pH set point as having statistically significant effects. In order for these results to be of value, the trends identified by the SimCell platform must correlate to those obtained in larger scale systems.

The results of this study demonstrated a good correlation between the microbioreactor and bench-scale platforms across 11 different process conditions. Based on this level of correlation, the SimCell platform and the concepts described in this study can be extended to larger experimental designs to quickly and efficiently explore a larger knowledge space and identify the design space in cell culture process development. 

Figure 2. Correlation of specific productivity between the SimCell platform and bench-scale bioreactors: An R2 of 0.90 was obtained for the 11 comparison runs between the two platforms.

Rachel Legmann, Ph.D., is principal scientist, and A. Peter Russo, Ph.D. ([email protected]), is product manager at Seahorse Bioscience. Web:

Previous articleLink Technologies Licenses University of Manchester’s Platform to Reduce Background Fluorescent Signal
Next articleReport Highlights Biotech Strengths, Opportunities, and Weaknesses in Newest EU Countries