Sue Pearson Ph.D. Freelance Writer GEN
Making the Challenge of Manufacturing Proteins a Commercial Reality
Since a large proportion of the cost of goods (COGs) of a biologic is determined during the drug design phase, it is critically important to design the molecule from a manufacturing point of view and compensate for heterogeneity right from the start.
That’s the view of Emma Harding, Ph.D., head of molecular chemistry, manufacturing, and controls (process research) at Glaxo-Smith Kline, who spoke at Knowledge Transfer Network’s recent Process Development for the Manufacturing of Challenging Proteins meeting.
However eliminating variance is not a simple task, as Andrew Kaja, senior scientist at GSK, pointed out.
“When producing biologics, cells are not predictable and don’t always generate product and express it in the supernatant at a specific time,” he explained.
Even when the protein is being expressed there can still be problems, which can occur at any stage of upstream and downstream bioprocessing.
“Proteins can ‘break bad’ by forming aggregates, precipitating out or co-purifying with undesirable proteins, all of which can affect the yields and efficiency of your biomanufacturing process,” added Richard Tran, Ph.D., principal scientist at GSK.
These issues can make manufacturing protein-based therapeutics within a commercially viable timeframe challenging but much can be overcome with considered approaches.
Finishing School for Biopharms
According to Dr Harding, GSK process research scientists perform a mini Quality by Design (QbD) assessment, known as a Molecular Design Intent (MDI) process on promising leads before molecules are taken through bioprocess scaleup. They use an in silico analysis to check for hot spots, including glycosylation and deamidation and to identify potential solubility and aggregation issues and predict how this might affect the pharmacodynamics, pharmacokinetics, and potency of the molecule.
Using this information, they rank their lead biological candidates as low, medium, or high risk for difficulty of manufacturing. When the candidates have been ranked, the scientists discuss if there are upstream or downstream process or genetic modification methods of eliminating those potential solubility and aggregation issues identified with the problem molecules.
“Our MDI 1 process is like a finishing school for biopharms, they have to pass this before we can consider investing time in moving them on into bioprocess development,” said Dr. Harding.
When molecules are being considered for scale-up, bioprocess scientists are faced with a number of options to optimize clone selection and process parameters to try to ensure that proteins will not present titer and formulation challenges. They can use the traditional shake flask and reusable benchtop bioreactors approach or rely on disposable versions of these to speed up the process. Increasingly, many are now utilizing fully automated micro bioreactor mimic systems.
“The scale-down equipment you use to predict process conditions will have an impact on the type, quality, and quantity of data you can produce, so you have to choose according to the complexity of information you need,” cautioned Kaja.
He then described how GSK is using a range of different technologies to solve its bioprocess scale-up questions. For reasons of simplicity, cost, and throughput, many initial screening projects use shake flasks and microtiter plates.
“When we want to further characterize clone performance in these uncontrolled environments we use the RAMOS automated shake flask system,” he noted. “For example, when we wanted to find out the effects of different maximum oxygen transfer rates and of changing concentration of IPTG for induction had on one set of clones, we used this system with cultures induced with either 10 µM IPTG or 500 µM IPTG and with fill volumes of 10 mL and 40 mL, respectively.”
From the automated independent gas analysis Kaja and his team were able to see that at the lower IPTG amount and in combination with a higher maximum oxygen transfer rate, the carbon source was depleted at 12 hours but at the higher amount of IPTG and with less oxygen available, the fermentation was maintained for longer and it was 25 hours before the carbon source was depleted.
“Using the shake flasks allows us to look at a couple of parameters and pick out strain specific differences. Then we can move promising clones to an HTS system, and this gives us a better and more robust screening technique,” he said.
Zeroing in on Costs and Time
For studies that require a larger design space than a shake flask, GSK is using the Pall Micro-24 MicroReactor system.
“It is too expensive and time consuming to run 24 cultures in benchtop bioreactors,” explains Kaja. “Using this system we can control and monitor pH, DO, and temperature in each well, and we’ve found using this system that titers are akin to those we achieve in fermenters, whereas titers are lower in the shake flask system so the Micro-24 is better at mimicking our fermenters in an HTS manner.”
Now the team uses this plate-based model for strain selection to rank host/vector combinations before fermentation development work takes place.
For the next stage in process development, GSK is replacing its stainless steel benchtop capabilities with 250 mL single-use benchtop bioreactors. This allows more focus to be placed on expanding and exploring a repertoire of process parameters and quality attributes.
“With the disposable bioreactor we can generate 60 models, which we layer and analyze to predict conditions for maximum titer and minimum aggregation during fermentation runs,” Kaja pointed out. “This helps us to determine the optimum time point to harvest the proteins so it is good for efficient process development.”
It seems that single-use benchtop bioreactors are becoming a popular choice for process development as, according to Adeline Fanni, upstream process development scientist at Actavis, they reduce set up, preparation, cleaning, and sterilization time from six to ten hours down to two hours.
“We’re looking at switching to single-use benchtop bioreactor because they come pre-assembled, there is less chance of the run failing due to incorrect set-up and since it is disposed of at the end of the experiment, this reduces the chance of cross contamination with material from a previous run,” Fanni explained.
However, one issue she cited with many single-use bioreactors is that they are closed and not versatile systems.
“There is not a wide choice of impellor or spargers and because of this many of these bioreactors will not perform well in perfusion high cell density fermentation where the oxygen transfer demand is high,” Fanni said.
For their process requirements, Actavis scientists evaluated the specifications of three 1 L single-use bioreactors and chose the Cercell bioreactor to run comparative tests against a standard glass benchtop bioreactor.
“We are using the Cercell CellVessel as it is customizable so we can have different impellors, spargers and sensors,” explained Fanni. “Our comparative results show that in a fed-batch culture of a mammalian cell line cultured in either the glass or single-use fermenters, the titer and cell density are comparable. We will continue to evaluate this single-use bioreactor for perfusion high cell density fermentation and will look to use this in process development to develop robust scale-up.”
Tweak the Molecule
During process development, there are some proteins that remain challenging even when the bioprocess parameters have been optimized. In these situations, Dr. Harding suggested: “changing the molecule slightly can have a much more positive outcome.”
She presented a number of case studies in which this approach proved successful. For example, she discussed a protein therapeutic that is soluble at pH 3 but increases in turbidity with increasing pH and at pH 7 (a pH that many protein therapies would be formulated in solution at) produces a very turbid solution.
“Trying to find a process solution for this is a huge amount of work so we used bioinformatics to determine if there were any areas where we could change the molecule to improve its hydrophobicity,” Dr. Harding explained.
By exchanging two tyrosine for two asparagines molecules in the protein, her team was able to produce double the titer and the molecule remained soluble at different pHs.
In a second case study, she detailed how a protein therapeutic cultured in a microbial fermentation formed a clumpy solution in fermentation when the culture reached confluence, making both culturing the cells and purification of the protein difficult.
Again using bioinformatics it was discovered that a single residue could be causing solubility issues so alternative single amino acid changes were exchanged for the residue. Noted Dr. Harding, “One of the swaps ensured the culture no longer expressed proteins that aggregated and the resulting fermentation produced 5 g/L of protein, which was easier to purify.”
This molecular tweaking is being taken to a new level by Edinburgh-based Ingenza, which provides services in synthetic biology, industrial biotechnology, and biologics production. It has developed a DNA construction platform known as inABLE® that uses oligonucleotide-driven gene assembly and other tools that enhance many aspects of recombinant protein synthesis.
“We use inABLE, a combinatorial genetics platform with designed oligonucleotide linkers, to build designer-genes for optimal protein production, making its expression simpler,” said Sam Capewell, Ph.D., senior scientist. “Initially, we divide the protein-coding gene sequence into sections and using these linkers we rejoin the protein in various predetermined combinations using microfluidics. In this way we can auto-assemble alternative combinations to produce a library of thousands of variants of the initial protein.”
Dr. Capewell presented a case study of the application of this technology, where one company had come to Ingenza with a key biologic target expressing truncated proteins resulting from inefficient and incomplete translation of the protein.
Using the inABLE platform, Ingenza scientists constructed a codon-usage variant library of this protein and with a lined gene fusion and colorimetric assay screened through almost 50,000 sequence variants. This led to the identification of efficiently expressed gene fusions, with verified enzyme activity and 50 optimal synthetic gene variant clones were selected for progress into development.
Dr. Capewell presented a second case study in which an academic collaborator had been having difficulty expressing a cyclic peptide that is cytotoxic to drug-resistant tumor cells. This is because the original protein is made in a native marine host and the amount the host could produce was too low to support development and functional characterization of the target molecule.
Again using their inABLE platform, Ingenza scientists combined clusters of cyclic peptide biosynthetic genes/orthologues and constructed E. coli strains that over-produced target cyclic peptides to provide sufficient amounts for pharmaceutical testing.
Commenting on the value of modifying the proteins, Dr. Harding concluded that “Even single amino acid changes in your target molecule can sometimes have a big impact and can mean you can manufacture a promising but difficult molecule that you would previously have abandoned.”
Speakers at the conference all agreed that producing protein-based therapeutics within a commercially viable timeframe can often be challenging. However, each presented how with considered approaches in upstream process development such as in silico analysis, using disposable technology and micro bioreactor models for process optimization, as well as smart protein engineering these issues can be resolved to ensure cost-effective manufacturing.