Pharmas find that development of protein therapeutics involves determining the perfect processes to produce unstable antibodies
JR Minkel
Minkel is a freelance writer based in Brooklyn, N.Y.
The “cryo-coma” in the clinical protein manufacturing facility of Wyeth BioPharma, Andover, Mass., is as big as a barn. Twenty 100-L stainless steel tanks sit in two rows. Only two of these cryo-vessels are in use, their activity given away by the black insulating jackets partially covering them. Each contains tens of liters of a frozen solution of experimental antibody that is bound for clinical trials. If the antibody requires some subtle change, the cost will be significant, because the manufacturing process is as frozen as the antibody itself.
The pharmaceutical industry is willing to put up with those headaches, however. “The market [for protein drugs] is growing very quickly,” says Robert Cyran, a reporter for Breaking Views, a financial publication, and a former analyst for Business Communications Co. Inc., Norwalk, Conn. “In general, it’s going to be growing much faster than small molecules.” Monoclonal antibodies for cancer therapy alone, which now constitute a several billion-dollar industry, could grow to $13 billion by 2008, according to a June report from Kalorama Information, New York.
Several criteria influence the decision to pursue a protein drug rather than a small molecule. The main requirement is that the target must be extracellular, either a cell surface receptor or soluble factor, because proteins are too large to enter a cell directly. If a missing or malfunctioning protein causes a disease, such as in hemophilia, a replacement protein is the obvious choice. Of late, the industry focused on antibodies, which act as antagonists and disrupt protein-protein interactions. In general, a protein drug is desirable if the target is not an enzyme, and if it has a large surface of interaction with one or more proteins, says Davinder Gill, PhD, director of antibody technologies, Wyeth Research. “Then a protein-based strategy makes sense and we would pursue it up front.”
Protein discovery follows the same basic stages as for small molecules, from target identification to lead optimization, says Sue Dillon, vice president of discovery research, Centocor Inc., Malvern, Pa. The characteristics of an optimal lead, she says, include a close match to the native human protein, high affinity, specificity, stability, and expression in cell culture. “In many ways, protein therapeutics are in their toddlerhood” compared to small molecules, says Steven Projan, vice president of protein technologies, Wyeth Research. “We’re starting to apply a lot of the lessons of small-molecule drug discovery,” he says, including structure-activity relationships. “But the beauty of proteins is because the binding surfaces involved are so large, we can get much more potent, much more specific interactions between a therapeutic protein and its target. We can also get much longer half-life molecules and, in general, much less toxic molecules.”
A major hurdle for protein drugs is delivery. They break down quickly in the stomach, and so must usually be injected. “What we need in the future for protein systems are novel delivery systems,” says Projan. “If someone’s got an interesting delivery system I’d love to talk to them.” Wyeth BioPharma collaborated with Medtronic Inc., Minneapolis, for example, to develop an absorbable collagen sponge for applying BMP2 to fractures during surgery. Researchers at Centocor, a division of Johnson & Johnson, are looking for ways to deliver compounds to the central nervous system and other compartments inaccessible by intravenous or subcutaneous injection, says Dillon.
Because of the delivery problem, and because protein drugs can’t yet act on targets inside cells, companies often pursue proteins and small molecules simultaneously. If a target is from a molecular family that researchers know responds to protein therapy, they might devote more resources to a protein therapeutic and follow up with a small molecule if needed, says Gill. If they have no prior experience with the target, he says, they will consider the available pathway information to decide. Centocor explores both small molecules and proteins for several classes of cell-surface receptors, including some G-protein-coupled receptors, says Dillon.
When to begin?
The next problem is when to begin production of a cell line. “The conclusion we’re coming to more and more is as soon as you know the protein that you think is a drug, you invest in the production cell line,” says Steven Arkinstall, head of worldwide discovery, Serono SA, Switzerland. “You’re doing [it at] a risk because your basic research carries on in parallel and you may decide later on that a different form of that protein is the form that you actually ultimately want to make into the drug, in which case the early investment in the first production cell line is wasted. But such is the pressure on us to produce quality production cell lines as quickly as possible that we would invest at the earliest stage possible. It’s such an intense process we can’t afford to cut any corners in terms of quality, and that takes time.”
Development is the stage at which proteins really distinguish themselves from small molecules. Compared to their organic chemical counterparts, proteins are extremely complicated and could potentially harbor “cryptic” elements capable of causing an immune reaction. So far, the only way to know for sure if a protein, even a humanized one, is immunogenic, is to test it clinically. “The consensus in the field is that there aren’t an adequate set of tests to fully characterize the molecule,” says Michael Kamarck, PhD, senior vice president, Wyeth BioPharma. “Therefore, we fall back on completely locking down the process by which the protein is made.” The clinical manufacturing process must perfectly match the commercial one for the drug to be acceptable for release, he says.
Development consists of establishing a small-scale production process that can be scaled up predictably. “The biggest hurdle is making sure you’ve generated a cell system that has the capabilities of making the protein in large enough quantities, that can be scaled up and produce the protein in a very high-quality manner, in a way that can be reproduced time and time again,” says Arkinstall. At Wyeth BioPharma, locking down the process begins once researchers select a clone transfected with a vector encoding the protein. Some of the frozen cells are put into a master cell bank in advance of the decision to proceed to development. All development from that point must refer to the master cell bank. “Once we establish it and lock it down, we’re stuck with it,” says Kamarck.
The process is difficult to unwind because testing a protein’s stability takes years, and production requires specialized manufacturing equipment. To change a purification step after securing regulatory approval, researchers would have to reexamine the antibody’s stability and have the new process approved again. The company builds commercial facilities during early phase II studies, he says, and the process has to be in place by the start of phase III trials or it’s too expensive to make changes. “That’s a pretty high-risk investment,” he says.
If a protein fails clinical testing, says Arkinstall, “you’d scrap [the production cell line] and you’d go with your backup. That’s part of the natural cost of R&D.”
Screening robot
The ClonePix robot, from Genetix Ltd., New Milton, UK, stands out immediately in Wyeth BioPharma’s Andover drug substance development lab. Sleek and black, its transparent hood resembles a futuristic, oversize barbecue. In place of a grill is a dark surface broken by five brightly lit cell culture plates, arrayed in an X. Dotting each plate are colonies of Chinese hamster ovary cells, each the product of a single cell transfected with a desired antibody. The task is now to find the clone that best expresses the antibody.
Scientist Kelvin Kerns is operating the robot. A motorized box crisscrosses the airless space above the plates, snapping their portraits with a CCD camera. Image processing software identifies and classifies the colonies, then projects the plate’s image onto a computer screen, where colored circles outline the colonies: blue means too small to sample, pink denotes colonies grown too close together, and green means just right. The robot is erratic this afternoon and doesn’t want to pick colonies—the lab has only had the machine for a few weeks—but at Kerns’s prodding the robotic arm starts moving and a thin needle jabs each selected colony and transfers it to a culture dish for assaying.
The robot reflects the protein development group’s goal of moving from transfection to production in 10 months or less. “We had to find tools to screen large numbers of colonies,” says Mark Leonard, director of cell and molecular sciences at Wyeth BioPharma. Picking clones manually took days; the robot takes a few hours. The group will whittle the clones down to a handful of candidates, and test these to identify the ones likely to grow best in the manufacturing facility’s bioreactors. Researchers will set up small-scale cultures of varying pH, temperature, and nutrient content to find optimal growth conditions. Along the way, they will assay the protein for target binding and possibly for glycosylation or other post-translational modifications.
“You’ve got to get it right because it’s very expensive to go back and do it again,” says Jeffrey Deetz, director of drug substance development, with Wyeth BioPharma. The group has begun using microrray experiments to figure out which genes create the most productive cell lines.
Once the cell line is chosen, the group screens protein purification conditions robotically. As each step of the process is optimized, the researchers scale up that step, culminating in tens of grams of the purified protein for further development.
Process development has benefited from a number of advances, says Deetz. Wyeth’s proprietary expression vector produces high levels of protein without an amplification step, a proprietary culture media minimizes metabolic wastes, and automation has made analytic support more efficient. Leonard works with discovery researchers to select the best antibody sequence for transfection. Sequences have certain properties that make them more likely to express well, fold and secrete correctly, garner the right post-translational modifications, and remain stable for longer times and soluble at higher concentrations. When the drug substance group is done, they deliver the protein to drug product development, which determines how to make it suitable for human administration.
Freeze-dried proteins
Nicholas Warne crunches across light snow on his way to the Andover Drug Product group, which he directs. In his hand he holds a 20-mL vial of white powder: a freeze-dried antibody. “It’s like freeze-dried coffee,” he says. “Once you get rid of the water, you can get very good stability. It would fall apart within days in solution.” The antibody itself is colorless, he says; the white color comes from sucrose, mannitol, glycine, and other stabilizers, which maintain the protein’s 3D configuration in the absence of water.
For that vial of protein, Warne’s department determined how to stabilize the protein for storage and human delivery, freeze-dry the protein, and construct the vial so it won’t interfere with the drug. Long-term stability testing takes years, so the group makes educated guesses. Formulators test various combinations of chemical stabilizers. Antioxidants prevent certain chemical reactions that erode proteins, buffers maintain the proper pH, and surfactants reduce the surface tension of the solution to minimize protein denaturation where air meets liquid. Calorimetry studies tell them the temperature at which a protein degrades.
One challenge of formulating antibodies is their relatively high concentration compared to secreted proteins, up to 100 mg/mL. As a protein’s concentration increases, it tends to clump together and come out of solution, which can cause immune reactions. Clumped proteins turn opalescent in solution, so Warne’s group performs light-scattering experiments to find the best conditions for solubility.
Researchers study the freeze-drying process by calorimetry and by lyophilizing small amounts of solution in a microscope. Lyophilization consists of freezing the solution and applying a vacuum to sublimate, or directly vaporize, the solid. If the vacuum occurs at the wrong temperature, the sublimation may take too long or it may turn the solution into amber goo. X-ray spectroscopy confirms that the large-scale process matches the small-scale one, and FTIR performed on the protein in solution and solid state ensures that its basic structural fold remains intact. The group also characterizes the freeze-dried stabilizers (the wrong conformation of glycin can damage the protein) and simulates freezing and thawing in small-scale cryo-vessels to troubleshoot that process.
Part of developing the production process means identifying tolerances in each stage and understanding how those tolerances interact. For example, lyophilization involves a freezing step and two drying steps, each of which has an optimal temperature and pressure. Drug product researchers study the effect on the protein of some critical combinations of temperature and pressure for each step. The result is a measure of the robustness of the process. “It’s like if you have a table,” says Warne. “You want to find the safest spot, which is right in the middle of the table, so if your production process goes off one side or the other you’ve got plenty of table to work on before it falls off.”
Finally, the process is essentially worked out. The cell line goes to the pilot lab, which produces hundreds of grams of product for animal toxicity testing and offers one more check that the process will scale up predictably. After pilot production the cells go to the development facility, where the process scales up another 100 to 500 times in order to supply Wyeth’s clinical trials. In these stages, engineers have to pay special attention to mixing and ventilation, which are different in a multi-hundred- or multi-thousand-liter bioreactor than on a bench top, and the process for removing cells from culture.
The culture is expanded in a seed reactor, typically for less than a week, to determine the right mixing times, pH, temperature, and gas flows. The reactor is sampled often to monitor cell count and protein expression. Once the process is completed, the culture goes to a multi-thousand-liter tank for two to three weeks, depending on the campaign and culture properties. Right now Wyeth’s protein trials are mostly phase I and II, which may require up to kilograms of protein. A 60% yield is typical after purification. For the facility’s largest reactor, which holds 6,000 liters, that means each liter has to yield a gram of protein. After purification, the protein goes to the cryo-vessels for freezing and finally to “fill and finish” for freeze-drying and placement into vials.
Controlling and monitoring the fermentation process is crucial. A locking-down process is built into the plant’s automation system, which allows reagent tanks to only be connected in particular ways, based on predetermined rules. Operators monitor the process through thousands of samples to make sure the cell culture has uniform and, therefore, reproducible cell densities, growth rates, and product titers.
Modular and flexible
The Andover facility is unusual, says Kamarck, in that it is highly modular and flexible. Notably, the central fluid distribution system can connect any two tanks in the building to match the volumes of culture and reagents needed. “In the very early days, we built a facility for a single molecule,” he says. “Everyone in the industry is moving toward trying to create more standardized approaches to use facilities we’ve already made in more general ways.”
Standardization and flexibility will be key in bringing down the price of protein drugs, says Kamarck. “If 60 to 80% of the products from biotech are antibodies, these products are likely to have lower costs in development,” he says. “It’s not a new molecule every time.” Improvements in process development should also help, he says, as will increased yields from cell lines and purification. “We are aware that these are expensive products. We’re not happy that they’re expensive products. I think we can make them less expensive to the patients as we get better at our job.”
Filed Under: Drug Discovery