A strategy for increasing process development workflow efficiency by incorporating enabling high throughput technologies systems including ambr™ bioreactors, the Cedex™ Bio HT Analyzer, and the Tecan Fluent™ pipetting robot.

During this year’s PepTalk conference, there was a very interesting presentation given by Dr. Timo Frensing, Large Molecule Research, Roche Pharma Research and Early Development, Roche Innovation Center Munich. The title of the talk was “Integration of ambr HT-bioreactor systems into the USP development workflow and into the data acquisition, management and analysis system.” In the talk, Dr. Frensing described how his team was able to increase process development workflow efficiency through the incorporation of several enabling high throughput technologies.

Dr. Frensing began his talk by providing background information about the Roche Pharma Research and Early Development (pRED) group. Roche pRED is one of three groups at Roche responsible for research and early development. His group in Munich has responsibility for discovery through process development and production of material up to Clinical Phase I. They are also responsible for transferring a successful process for manufacturing Phase I material.

Drivers for increasing upstream process development workflow efficiency

Dr. Frensing explained that his pRED recognized that their upstream process development workflow was inefficient. It was highly manual and the use of 2L bioreactors didn’t permit a high throughput process, so they were not able to run as many fermentations per run as they wanted. They recognized that their workflow could become more efficient by running multiple fermentations at the same time with the use of automated bioreactors. For instance, in their current system samples were collected manually from the 2L bioreactors and then prepared and sent for pH, and cell counting. Samples had to be centrifuged to generate cell free samples that were then sent to their Cedex Bio HT for metabolic and titer measurement. Samples also had to be prepared and sent for osmolarity measurement. Lastly, backup samples were collected and stored. These were all manual processes that consumed a large amount of operator time and resulted in a process was low throughput, time consuming and permitted only a small number of samples and thus a smaller data set.

Increasing process development workflow throughput by incorporating automated bioreactors

To increase throughput and improve efficiency, the team decided to incorporate automated bioreactors into their process development workflow. In order to determine their suitability against current operations (2L bioreactors), Dr. Frensing and his team tested Sartorius’ ambr250 and the ambr15 to see how they compared with the 2L glass bioreactor results. In the comparison study, the three systems were run in parallel for one year on every project. The team was primarily interested in evaluating the systems for final clone selection. During the year, 42 CHO lines were evaluated for 5 different antibody formats. The formats chosen were very complex protein candidates, including a brain shuttle antibody fusion and a T cell bi-specific antibody. Inoculation was done with the same pre-culture to ensure a direct comparison.

Results of the comparison study

During his talk, Dr. Frensing shared the system comparison data for a final clone screen of 8 clones. The team found that the three systems were comparable in evaluating growth and titer with very minor deviations. However, when they looked at specific parameters, they did find differences between the systems. When looking at ammonium, only one clone showed comparable results. The ambr15 culture had particularly high levels of ammonium and there was also variation among the systems for lactate and viability results.

In order to conduct a thorough analysis of the data, the team asked their bioimformatics group to conduct a detailed statistical analysis. They found that in order to fully compare the results, they needed to employ both the two one-sided t-test (TOST) and the model comparison method (Islam et al., 2007). Using these two methods in combination, a statistical analysis was generated that compared each of the ambr methods to the 2L glass bioreactors. Results showed that the ambr250 was slightly better across all parameters, but with the most critical parameters, the ambr250 was much better. The team then evaluated comparability for product quality and all systems were comparable. With glycosylation the three systems were mostly comparable, however there were some deviations with high mannose structures in the ambr15, which Dr. Frensing said, could be do to the higher ammonium levels in those cultures. In clone ranking, with titer as the most weighted value, the ambr250 had comparable results, but there were some differences with the ambr15 system.

After all the data was in, they conducted a system comparison where they ranked each of the systems high, medium and low for a set of criteria in order to determine appropriate applications for each system. The criteria and associated rankings are listed below:

• Data match and clone ranking compared to the 2L glass bioreactors – ambr250 was ranked good and the ambr15 was ranked satisfactory.
• Instrument reliability – the 2L bioreactors were ranked medium, ambr250 was ranked high and the ambr15 was ranked low. The ambr15 ranking was due to a failure of some fermentation runs, this meant that 24-48 fermentations were lost each time a run failed.
• Lab space and cleaning time – the 2L bioreactor being a poor choice, ambr15 being good and ambr250 medium.
• Cost per fermentation – 2L bioreactors had a medium cost, ambr15 had a low cost and ambr250 had high cost, however Dr. Frensing clarified that when you consider labor, set up and cleaning costs, the ambr250 was less expensive than the 2L bioreactors.
• Throughput – low for 2L bioreactors, high for the ambr15 and medium for the ambr250.
• Flexibility – the 2L bioreactors offered the most flexibility in terms of being able to test different stirrers, spargers, sensors, etc. They also provided the highest sample volume, which allowed more characterization work to be done. The ambr15 offered the least flexibility and the ambr250 offered medium flexibility in that you had enough volume to conduct some additional sample characterization based on the higher sample volume.

Based on these considerations and how each system performed, Roche pRED decided to utilize the 2L glass bioreactors for small work packages and in testing new equipment, such as probes and sensors. They decided to use the ambr15 for early clone screening and process development due to its high throughput and low cost. They decided that the ambr250 would be best for their final clone screen and for process optimization.

Implementing the improved process development workflow

As the team incorporated ambr15 and ambr250 bioreactors into the process development workflow, the number of samples increased. The large number of samples required high throughput sample processing and analytics. To accomplish this, they purchased the Tecan Fluent pipetting robot and used it to link the ambr systems and the Roche Custom Biotech’s Cedex Bio HT Analyzer.

Key benefits identified by Dr. Frensing of using the Tecan Fluent to link ambr systems with the Cedex Bio HT Analyzer:

• Fast and automated workflow
• 12 – 192 samples can be processed at once
• 48 samples can be processed within 20 minutes (manual processing: 60 minutes)
• Operator only has to spend ~5 minutes hands-on
• Higher and operator-independent accuracy for sample pre-dilution

The next challenge was to take the large amount of data being generated with this high sample, high throughput process and create a high throughput data processing workflow. Dr. Frensing describes that they were able to collect the online data via the PI Processbook and send it to the data warehouse. From there, they could use the data for visualization, evaluation and reporting. They ran their offline data through a central management tool, which collected the data into an electronic notebook. However, they now also needed a middle software to communicate between sample management and the analytical devices. For this, they chose the Smartline D@ta Cockpit software. With Smartline D@ta Cockpit, they were able to collect data and send it to the sample management tool and then transfer to the data warehouse. Sample tracking and registration had been done manually, but with the new systems, it was too much work. Plus they were manually accessing sample data to be able to adjust feeding regimes in the ambr systems. They wanted to automate both of these functions, so they asked the creators of Smartline D@ta Cockpit to expand the software function to permit sample tracking and registration using the software and to be able to automatically send the information to the analytical machines. This expansion has been done in development, but has not yet been launched because they are working on also developing complete sample tracking through the workflow, which will allow data to be sent back to the ambr systems where sample data results, such as glucose levels, could be used to automatically adjust feeding regimes.

Summary

In summary, Dr. Frensing stated that the analysis of the three systems led the team to select the ambr15 system for clone screening and process development and the ambr250 system for final clone selection and process optimization. They were able to establish an efficient link between the ambr systems and the Cedex Bio HT Analyzer using a Tecan Fluent system, which permitted them to improve their process development workflow, make it semi-automated, and significantly more efficient. They redesigned their data management platform to handle the large number of samples and data being generated by automating online and offline data transfer to the data warehouse. They are also developing an automated data feedback loop between the ambr systems, the Tecan Fluent and the Cedex Bio HT Analyzer.

Recommended Further Reading:

Automated Collection and Analysis of Bioreactor Samples to Enable Quality by Design Initiatives