Send to printer »

Feature Articles : Apr 1, 2012 (Vol. 32, No. 7)

Small-Scale Modeling and Predicting

  • Vicki Glaser

Optimizing production efficiency of a bioprocess to yield high titers of a desired recombinant protein or monoclonal antibody requires great attention to detail. This begins at initial design and testing of a small-scale process right on through scale-up to industrial-scale manufacturing.

At every step decisions should be based on a thorough understanding and characterization of the process components (cell and media selection), process parameters (affecting product quality and quantity; defining an acceptable range), and the target product (identifying a product profile that ensures maximum safety and efficacy).

Achieving these goals demands the integration of available technology and analytical tools with a clearly defined quality by design (QbD) strategy, in which quality and reproducibility are built into the process from the outset, using scale-down process design and analysis methods that reliably model the environmental factors affecting cell growth and viability and protein production in a pilot- or large-scale bioreactor.

At the BioProcess International Europe Conference in Prague later this month, Daryl Fernandes, DPhil, founder and CEO of U.K.-based Ludger, will discuss a range of topics related to drug design, manufacturing, and biopharmaceutical comparability.

There is no question that building QbD into drug development can be time and labor intensive, the payoff comes in having a well-characterized, well-understood process for producing a drug compound efficiently and reproducibly at small scale that can be reliably and predictably scaled up to manufacture commercial quantities of a comparable drug that is safe and effective and will pass regulatory scrutiny.

Key Principles

Dr. Fernandes will focus on key principles of QbD—determination and prioritization of critical quality attributes (CQAs), development of an optimized quality target product profile (QTPP), design of experiments (DoE), design space, and control spaces. These concepts will be illustrated primarily by case studies for drug glycosylation patterns.

Glycosylation is a highly important structural feature of biopharmaceuticals, but it can be challenging for both biopharma companies and regulators. It can significantly affect safety and efficacy profiles of glycoprotein therapeutics, but also add greatly to their complexity, heterogeneity, and variability. It is one of the most difficult factors to control in biomanufacturing.

In particular, it is not uncommon for the glycosylation pattern to change during scale-up, sometimes leading to serious regulatory issues.

Dr. Fernandes points to the A-MAb case study, prepared by a consortium of companies known as the CMC-Biotech Working Group, as a good example of what can be learned from the implementation of QbD and of the importance of glycosylation factors. Nearly half of the potential CQAs used to model the design space in the A-MAb study were glycosylation parameters.

“A-MAb was a superb study. It showed how QbD can help drug companies establish well-controlled, well-characterized biomanufacturing processes. But it also revealed some difficulties with implementing the principles in ICH Q8/Q9. These include the high cost of DoE studies, the troubles when using FMEA to determine and prioritize potential CQAs, and the feasibility of obtaining an expanded design space,” says Dr. Fernandes.

The mistake made most often in applying QbD to glycoprotein therapeutics has been “starting expensive, time-consuming DoE with an un-optimized QTPP,” he says. Instead, he recommends that companies “invest the time to develop a well-thought-out QTPP with structural features, including glycosylation patterns, designed to optimize the drug’s clinical performance.”

One can then select an expression system and parameters to deliver a therapeutic that matches that QTPP as closely as possible.

Controversially, Dr. Fernandes argues that an ICH-type design space is not an essential part of QbD, and that a carefully designed QTPP together with a conventional biomanufacturing process control system should be considered as a first level of QbD implementation.

However, he also emphasizes that using design and control spaces could significantly improve the consistency of batches of glycoprotein therapeutics and that following a QTPP with optimized, simplified glycosylation can reduce the DoE work needed to develop these.

Developing an optimized QTPP together with design and control spaces is particularly useful for reducing the problems of altered glycosylation during scale-up. Instead of taking the traditional route to scale-up and designing and characterizing a process at small scale and then assuming that by monitoring and controlling those defined parameters through scale-up the product will not change substantially, he emphasizes another approach. He cites the importance of devoting the time and resources necessary at the DoE stage to understanding what the environment around each cell will be like in a large-scale bioreactor.

Companies should then use a scale-down modeling approach to vary and mimic those conditions at small scale, defining the optimal design space and acceptable ranges for key CQAs, and use that model as the basis for scale-up. With this strategy, scale-up should yield a more consistent cellular environment and be more predictive of what will take place in the bioreactor as process scale increases.

“The value of QbD comes in advancing your understanding of your drug and characterizing your process, not necessarily in having an expanded design space,” says Dr. Fernandes. For glycosylation, optimizing the glycoform population of a drug by increasing the percentage of high-activity isoforms can enhance its safety and efficacy profiles and simplify the drug profile, making it easier to scale up.

Dr. Fernandes adds, “One of the biggest advantages of QbD is in bringing people together early on in a drug development project, from across departments and disciplines, to develop a shared model for systematic realization of that drug—i.e., what it should do for the patients, how it should work in vivo, what it should look like (structurally, at the molecular level), how it should be characterized, and how it could be manufactured.”

Developing Predictive Tools

Martin Jordan, Ph.D., a scientist in the Upstream Processing, Biotech Process Sciences division of Merck Serono, will discuss work aimed at establishing a 96-deep-well plate platform for fed-batch processes for use in cell line screening and process development.

Dr. Jordan and colleagues transferred an entire fed-batch process that was being carried out in tubes or flasks on a shaker platform to 96-deep-well plates, keeping the physical conditions to which the cells were exposed as close as possible to the ones present in a bioreactor.

Enabling this work was what Dr. Jordan describes as “an optimal combination of technologies available for the 96-well-plate format”: a robotic liquid-handling platform for automating processing steps such as media preparation, seeding, sampling, and dilutions; a deep-well culture system with controlled oxygen and carbon dioxide levels; and high-throughput analytical tools for tasks such as cell counting and titer assessment.

He identifies the main bottleneck at present as the relatively slow turnaround time of the robotic liquid-handling system. For complex tests with variable doses for each well, process times are about two hours per plate.

“For large tests run in parallel on multiple plates, protocols need careful adoption regarding time-critical steps,” he says.

At Merck, cell line selection studies begin as soon as stable cell lines are established and ready for transfer from static to shaken plates.

“We adapt cells for two weeks (about five passages) to the growth conditions in suspension before launching our evaluation in the generic fed-batch process,” explains Dr. Jordan. “Currently we can do this for up to 500 cell lines in parallel.”

At every passage the researchers check each well for cell growth and adjust the dilution as needed. Throughout the process they use measurements of cell number and titer as indicators of growth, maximal cell density, viability profile, integral of viable cells (IVC), and productivity.

Dr. Jordan reports that initial proof-of-concept studies have shown this strategy to be “an excellent screening tool, since a large number of cell lines can be early tested under specific process conditions. Only a few top candidates will finally be evaluated in a bioreactor.”

The group has completed proof-of-concept with a limited number of cell lines.

“As a next step we will do a full study with the maximal number of cell lines in parallel with the current cell line generation program for at least one project.”

Host Cell Proteins

Late last year, at the BioProcess International conference in Long Beach, CA, a team of scientists from Life Technologies demonstrated the potential for developing a profile of secreted host cell proteins (HCPs) and using a subset of those proteins as an indicator of the level of recombinant protein being produced by the cells.

By understanding how changes in the signature of the host cell secretome relate to product titer, it may be possible to use this proteomic strategy for evaluating clones upstream of selecting a high-producing production cell line, for media development and optimization, for assessing cell health and viability, and for process monitoring.

The researchers described the use of quantitative mass spectrometry to measure changes in the levels of secreted HCPs in transfected cell pools that exhibited low, medium, or high levels of recombinant antibody expression.

The cells were grown in chemically defined, protein-free media so the HCPs would be easily identifiable. The team identified distinct patterns of secreted HCPs in the cell pools that correlated to the varying degrees of antibody expression.

They developed HCP fingerprints that comprised a subset of 50 proteins from the total of 450 HCPs identified on mass spectrometry. This subset showed the most dramatic concentration changes associated with variability in antibody production.

Moving forward the researchers will refine the HCP signature related to protein titer and also try to develop a secreted HCP signature that correlated with the quality of recombinant protein produced. While the results presented in the poster are specific to the model studied, Peter G. Slade, Ph.D., staff scientist, molecular and cell biology R&D at Life Technologies, believes that the concept can be generalized and this strategy for nondisruptive monitoring of the secreted proteome can have multiple applications.

“The model represents another tool available for people who want to get an indication of the general health of their culture and to understand how the protein profile links to other quality attributes of the culture,” says Stephen Gorfien, Ph.D., senior director, bioproduction R&D.

Parallel Processing

Earlier this year, Dasgip, a developer, manufacturer, and supplier of parallel bioreactor systems and accompanying bioprocess control software, was acquired by Eppendorf, a combination that Thomas Drescher, Ph.D., Dasgip’s CEO, calls “an excellent fit.”

As a result of the acquisition, Eppendorf gained Dasgip’s expertise in the area of parallel bioprocessing, complementing the bioreactor product line it gained with the earlier acquisition of New Brunswick Scientific. Dasgip benefits from access to Eppendorf’s range of products and expertise in the areas of liquid, sample, and cell handling and to its global sales and service capabilities.

Eppendorf now offers products for bioprocess research applications, development, and scale-up ranging from 30 mL to 3,000 liters.

Falk Schneider, Ph.D., head of software development at Dasgip, describes a key trend driving the bioreactor market over the past 6–12 months. “We are recognizing an increasing demand for cultivation in a scale below 1 L in the market,” he says. “This demand is driven by applications in the fields of strain and cell line screening, media optimization, and process development.”

Furthermore, he notes, “QbD methods have gained a broad awareness over the last year and are considered to be a tool to minimize the necessary amount of cultivation runs and at the same time gain the most valuable information out of them.”

Dasgip recently introduced the DASbox mini bioreactor system. It is a controlled stirred tank bioreactor with a working volume of 60–250 mL. Drawing on Dasgip’s hardware and software design innovations for parallel bioprocessing, the DASbox comes in units of four, allowing for 32+ fold parallel systems to operate simultaneously.

A 24-fold system consumes only six feet of lab bench space, according to Matthias Arnold, Ph.D., head of hardware development.

“The burden of liquid cooling and heating has been completely eliminated by using peltier elements to control the temperature of bioreactors as well as condensors.”

An overhead drive covers the full range from 30 rpm to 3,000 rpm and can serve cell culture as well as microbial applications at the same time. The DASbox also incorporates mass flow controlled gassing.