Accelerating Data Access in Bioprocesses with Automated Aseptic Sampling

A typical perfusion bioreactor run requires approximately eight hours of labor for sample acquisition and at-line analysis, oftentimes requiring operators to be available during weekends and off-hours. During process development, when many conditions are run in parallel, the sampling and analysis workload can increase dramatically. This limits the potential output from a set of experiments, either by capping the maximum number of conditions that can be tested at once or through less frequent access to data. Additionally, results from off-line analysis of samples can take weeks to be returned. For these reasons, experimental results are often less complete, robust, or not as timely as desired; the resulting lack of information can then lead to incorrect decisions or wasted time during process development.

To address these challenges, the MAST^® Autosampling Solution was installed to automate aseptic sampling and sample management. The system allows sampling from ten points (i.e., bioreactors) and sample delivery to four primary analyzers. With this automated system, sample source sterility can be maintained while improving process understanding with near real-time and accurate analytical results.

Proof-of-concept Study: Integration of PAT Tools and Automated Sampling in Upstream Bioprocesses

Set-up of Experiments

An automated sampling system, MAST^® Autosampling Solution, was connected to the bioreactor to collect samples automatically for on-line analysis. In parallel, a Raman analyzer, ProCellics™ Raman Analyzer, was inserted into the head plate of the bioreactor for continuous in-line measurements. A 3 L stirred tank bioreactor was operated in perfusion mode utilizing a cell retention device. Fresh medium was continuously fed into the bioreactor, while spent medium and product harvest were continuously removed using a cell retention device. The MAST^® Autosampling Solution was integrated into the bioprocess workflow and connected to an on-line analyzer; both on-line and in-line data were jointly processed by the software.

Evaluation of Cell Culture Sterility and Sample Accuracy

Following installation, a proof-of-concept study was performed to evaluate the ability of the MAST^® Autosampling Solution to maintain sterility of the cell culture and accuracy of the sample data.

A set of three perfusion runs was performed:

• Two in steady state perfusion mode at 80 × 10⁶ cells/mL for approximately 16 days
• One in N-1 perfusion mode in which the cells grew to a density of about 150 × 10⁶ cells/mL to challenge the system and ensure that at higher densities with higher viscosities, the system could still measure the samples effectively.

Over the course of the three runs, the MAST^® Autosampling Solution took samples four times a day or every six hours. Once a day, a manual sample was taken within ten minutes of one of the automated samples to assess accuracy of the data between the automated and manual samples.

As shown in Figure 1 A, the desired level of cell growth was achieved across all three runs. No decline in VCD over time and a constantly high viability level confirms no contamination, indicating that sterility of the system was maintained for both sampling methods. At the high cell density, the MAST^® Autosampling Solution was able to deliver cell culture to the destination or analyze the samples. When comparing the automated sample data to the manual sample data taken once a day, across a variety of analytes, there was good correlation as indicated by results falling close to the y = x line (Figure 1 B).

Benefits of Automated Sampling in Perfusion Process Development and Method Robustness

The benefits of automated sampling in perfusion process development were studied. Runs were performed in a 3 L glass bioreactor with a 2.2 L working volume in steady state perfusion mode with a target cell density of 80 × 10⁶ cells/mL. Cell health was maintained with a constant perfusion rate of 20 pL/cell/day, or approximately 1.5 vessel volumes per day, using an EX-CELL^® Advanced HD Perfusion medium.

Manual samples were taken daily, as per normal protocol, which required about eight hours of operator time including weekend work. Additionally, samples were sent to an off-line lab for analysis of product quality, which resulted in a six-week turnaround time. In one of these process development experiments, the MAST^® Autosampling Solution was implemented and connected to a BioProfile^® FLEX2 (Nova Biomedical Corp., Waltham, MA, USA) for measurement of nutrients, metabolites, cell density, osmolarity, pH, and gases.

Figure 2 shows the resulting data for glucose, viability, ammonium, and lactate, as compared to the daily manual sample data, indicating similar trends regardless of the method. The use of automated sampling provided an increase in the frequency of data acquisition with much less effort; four times more data were collected, and the frequency could have been increased even further. The only resources needed for analysis of the automated samples was a 20-minute setup procedure and periodic cleaning procedures as initiated by the operator, eliminating eight hours of time spent on sample acquisition and analysis in a manual run. Additionally, the increased frequency of data allowed for improved comparison between conditions across other process development runs and improved the decision-making capabilities of the team.

The robustness of the MAST^® Autosampling Solution was also assessed across a set of process development runs. Three different bioreactors were run in parallel with 171 total samples scheduled. Only five samples were missed, providing a 97% successful completion rate and meeting the expectations of the team. Reason for the five (out of 171) missed samples was either a clog within the analytical instrument, the instrument being out of calibration or a lack of reagents needed for the analysis. In a standard production setting, these issues would have been easily prevented by regular maintenance. However, it's important to note that this proof-of-concept study was conducted at a smaller scale and did not implement the typical maintenance procedures found in a production environment. Therefore, these 5 missed samples do not significantly impact the findings.

Opportunities of a Combined Approach

As noted above, automated sampling can supplement other PAT tools for reduced effort and improved results.

The accuracy of the MAST^® Autosampling Solution was confirmed by in-line Raman measurements. Depending on the process, this allows biopharmaceutical manufacturers to choose between in-line real-time methods and the equivalent on-line measurement approach using MAST^® Autosampling Solution - whichever is more suitable for the individual situation. In addition, it opens up exciting opportunities for PAT method and model development utilizing two complementing technologies. Relying on the combined in-line Raman and on-line MAST^® Autosampling Solution approach empowers operators to develop robust and cross-validated measurements, enabling optimal process understanding and control.

Figure 3 shows an experiment in which automated sampling was used together with the ProCellics™ Raman Analyzer.

Following development of a chemometric model for N-1 perfusion cell cultures, data were monitored in-line to assess transferability to dynamic perfusion cultures. Results of the study showed that the in-line Raman sensor data was comparable to automated sampling on-line data with the MAST^® Autosampling Solution. The automated sampling data could be used to validate this model and pick up on deviations if they had occurred. Additionally, when implemented in the model development phase, automated sampling significantly reduced the workload to gather sample data. This approach can lead to shorter model development timelines with fewer replicates needed to obtain the same number of samples, and improved robustness by capturing the full scope of process variability with more frequent samples.

Conclusion

These experiments describe the initial steps towards obtaining on-line data from a bioprocess, demonstrating the value in an upstream process development lab. Results of the presented proof-of-concept study confirm the accuracy and robustness of the automated sampling approach using MAST^® Autosampling Solution in perfusion cell culture, while effectively maintaining sample source sterility. Implementation of automated sampling helps alleviate the burden created by the need for manual bioreactor sampling, providing improved data sets for quicker and more effective decision-making. Automated sampling fits an industry need to acquire on-line sample data and can be used in combination with other PAT tools for optimal results.