|  |
Glycosylation Matters
Faced with a looming shortage of capacity to produce recombinant protein-based drugs, biotech companies are busily exploring their options. And it's not just about lining up a contract manufacturer or committing $200 million to $400 million to construct a plant. Maximizing product yield is important, too, as is choosing the right cell line or host production system. There are lots of choices: Although mammalian cell lines are generally used to make recombinant proteins, bacteria, yeast or insect cells might prove more suitable. Some adventurous companies are investigating the feasibility of using transgenic plants or animals for large-scale production.
Unfortunately, no choice is clear-cut. Each has advantages; each has drawbacks. But if you talk to researchers who make their living tweaking protein production systems to maximize product activity and yield, they'll no doubt bring up the issue of glycosylation, irrespective of the production system used. If the recombinant protein doesn't have the correct sugar groups attached to its backbone -- in the proper locations and appropriately linked -- there's a good chance that protein will perform poorly in humans. It may have a low therapeutic efficacy; it may be cleared from the body too quickly; or it may even be immunogenic.
However, improper glycosylation might not always be a problem. In some cases, it might even be advantageous to have an immunogenic glycoprotein -- as the basis of a vaccine, for instance, where the idea is to stimulate the immune system. In other cases, it might not matter at all whether the glycoprotein has the correct sugar groups -- if the product is to be given topically, for instance, rather than systemically, the body's immune system won't enter into the equation.
Moreover, glycosylation isn't always necessary for therapeutic efficacy. Indeed, some recombinant proteins on the market aren't glycosylated at all, including Chiron Corp.'s Proleukin (recombinant interleukin-2) and Amgen Inc.'s Kineret (recombinant interleukin-1 receptor antagonist) and Neupogen (recombinant granulocyte-colony stimulating factor). But most are, including Amgen's Epogen (recombinant erythropoietin) and Aranesp (second-generation Epogen), Genentech Inc.'s Activase (recombinant tissue plasminogen activator) and all monoclonal antibodies.
Fortunately, glycosylated proteins don't have to be 100 percent identical to their natural counterparts to demonstrate efficacy: Reportedly, MedImmune Inc.'s humanized monoclonal Synagis consists of seven or eight glycoforms. Obviously, the FDA finds a certain degree of microheterogeneity acceptable.
So, when it comes to picking a production system, if the protein of interest isn't glycosylated at all, the choice is easy: Use Escherichia coli, molecular biology's workhorse. (Proleukin, Kineret and Neupogen are all made in E. coli.). But if a glycoprotein is at stake, that's another matter. All other production hosts -- yeast, insect cells, transgenic plants, transgenic animals (including chicken eggs) and even Chinese hamster ovary cell lines (CHO) -- add sugars, but each in its own way. And none are human-identical.
"The CHO cell line was selected in the beginning as the standard mammalian expression system because it is able to glycosylate proteins in a manner very similar to the way most human cells glycosylate," explained David Zopf, executive VP at Neose Technologies Inc. "But it is not identical. CHO cells do a reasonably good job of reproducing [human] glycan profiles, and they don't attach any sugar groups that human cells never use," meaning that the recombinant glycoproteins are not likely to be immunogenic. However, he continued, CHO cells don't necessarily complete the entire glycan structure (a complex carbohydrate that consists of branched chains containing several sugar moieties; see the illustration below). "When scientists are pushing culture conditions hard to maximize yield, the cells frequently don't keep up with the glycosylation, leading to shorter chains."
Insect cells, which can be grown under conditions similar to those used for CHO cells, "can be more efficient [than CHO cells] in the amount of protein they make per liter for some proteins," he said. Moreover, "they generally put sugar chains in the same positions on the proteins as mammalian cells do (although occasionally they choose a different spot), but the glycan structures are significantly different. Only the region that's closest to the protein is preserved; the outer parts of the glycan structures are missing, resulting in very short chains," Zopf continued. "Sometimes the protein can't fold correctly or it's less stable. It's susceptible to protease degradation, and it could be immunogenic."
Graphic illustration courtesy of Neose Technologies Inc.
Yeast, which also mimic mammalian cells to a limited extent, confound the issue by introducing different linkages. "Yeast put on N-linked glycan chains [which are linked to the amino acid asparagine] the same as CHO and human cells, but yeast also add O-linked glycan chains [which are linked to the amino acids serine or threonine] in places that CHO cells never do, and they use inappropriate O-glycans," he explained. Moreover, yeast N-glycans consist entirely of mannose, and the resulting structures are very large. High mannose chains can be immunogenic, and proteins containing them are rapidly taken up by the liver, limiting the drug's bioavailability.
One way around this problem is to modify the yeast so they don't synthesize these large mannose structures, Zopf said, "but most yeast don't like that. They grow slowly and they're not robust." Still, it is possible to use these organisms to successfully make a recombinant protein -- GM-CSF (Leukine, which Immunex Corp. just sold to Schering AG) is a good example.
Neose Technologies has developed another approach to solving the sugar problem: The company's GlycoAdvance technology uses enzymes to complete the carbohydrate structures on glycoproteins after they have been secreted from CHO cells. "We can modify the glycoprotein after it's been secreted," Zopf explained, "We use enzymes to remove the offending sugar" and then replace it with a naturally occurring one. "We can pare back the sugars to the core structure and then selectively build out glycoforms and test them," he added. Neose has got the technology worked out for transgenic animal-made proteins, too, and is developing the particular enzymes necessary to correct the glycosylation patterns on recombinant proteins made in yeast, transgenic plants and other host systems.
You might think that transgenic animals would be able to glycosylate properly -- at least as well as hamster cell lines. And, "in general they make glycan structures that fall into the same family as human ones. However, they are significantly modified in certain ways. Rabbits and larger animals terminate N-glycan chains with a different sugar [than humans use] or they use a sugar that humans make but attach it in a place that humans never do."
According to Zopf, this can result in "the same immunological problems that occur when you try to transplant an animal organ into a human. The hyperacute rejection of pig organs is a response to carbohydrates." In particular, it's the alpha 1,3-galactose linkage that presents such a problem, creating an antigen that is recognized as foreign. (For more information on xenotransplantation, and research efforts to overcome the rejection problem, see the Signals article, "The Transplantation Xeno-Derby.")
Other transgenic animals, he continued, will add N-glycolylneuraminic acid as the terminal sugar (instead of N-acetylneuraminic acid), and that small difference "is enough for the glycoprotein to be recognized as foreign," he said. So far transgenic companies haven't seen an immune reaction in humans, Zopf continued, but there is a lingering concern. Years ago when antibody-containing horse serum was used for passive immunotherapy, patients who received a second shot developed serum sickness [which is very similar to a hyperimmune reaction]. "A major component of that response was against N-glycolylneuraminic acid."
"The glycosylation patterns we get [in transgenic animals] are no more different from the native molecule than those produced in CHO cells," explained Tom Newberry, spokesman for GTC Biotherapeutics Inc. (formerly Genzyme Transgenics Corp.). Along with a number of other potential therapeutics under development (many of them for partners), the company's using transgenic goats to produce recombinant human antithrombin III (rhATIII), a blood plasma protein with anti-coagulation and antiinflammatory properties. It's taken that product into humans, too, and is currently conducting a pharmacokinetic study in patients who have a hereditary deficiency of ATIII. In the clinic, rhATIII "has a slightly shorter half life [than the natural protein] but a slightly higher binding affinity. Therefore, the clinical impact is negligible," Newberry said.
Added Daniel Coutos, GTC Biotherapeutics' director of process development and engineering, "The majority of our products don't show much difference in pharmacokinetic studies. They're cleared [from the circulation] at the same rate as proteins produced in CHO cells." If GTC Biotherapeutics does run into a glycosylation problem with a transgenic protein, Newberry continued, there are two ways to address it -- at the molecular biology level or via post-translational modification.
Transgenic chickens, apparently, use N-acetylneuraminic acid to glycosylate their proteins -- like humans -- meaning that therapeutic proteins made in chicken eggs are less likely to be immunogenic. According to Michael West, president and CEO of Advanced Cell Technology Inc., "The chicken is in many respects the most human of all the animal transgenic models." (For more information on progress in the use of transgenic chickens to produce therapeutic proteins, see the Signals feature, "Transgenic Eggs: A Golden Opportunity?")
While the glycans on recombinant proteins made in transgenic goat milk and chicken eggs appear to be fairly similar to their human counterparts, transgenic plants "do things to sugar chains that are superficially similar but they are modified in fundamental ways," according to Neose Technologies' Zopf. "For one thing, the chains are shorter. For another, plants use sugars and linkages that humans never do." For instance, the fucose linkage is different (see the illustration), and "it's recognized as a foreign antigen." And plants use xylose (also immunogenic), which is never linked to a sugar chain in humans," he explained.
But that might not matter, especially if the plant-produced proteins will be used topically. Clinical trials have already been conducted on an anti-Streptococcus mutans secretory monoclonal antibody generated in transgenic tobacco plants. Given topically, this product completely prevented re-colonization of human teeth with S. mutans (which causes dental cavities) for at least four months. According to Mich Hein, president of transgenic plant company Epicyte Pharmaceutical Inc., these trials demonstrated that the antibody was both safe and efficacious.
Other companies have injected plant-made proteins into humans in clinical trials, including a corn-made recombinant human gastric lipase for treating exocrine pancreatic insufficiency and tobacco-produced, patient-specific cancer vaccines. So far, no problems have been reported. (For a detailed description of these products, as well as an overview of protein production via transgenic plants, see the Signals feature, "Molecular Farming's Factories.")
"When we look at homologous systems, we see no deleterious effects on the protein's circulating half-life," Hein said. "If you produce a mouse antibody in a plant and then inject it back into the mouse, it behaves just like a mouse antibody made in cell culture." And, he added, "If there are any problems, there are straightforward ways to deal with them," including modifying the glycans post-translationally or even genetically engineering plant cells so they make different glycans.
In the future, it's possible that companies will choose a production system based on the characteristics of the particular protein they wish to make.
Non-glycosylated proteins can be made in bacteria. Secreted glycoproteins might be best produced by cultured mammalian cells (especially CHO cells), or perhaps insect cells.
Transgenic animals might be the best host for cell-bound proteins, which apparently are difficult to express in bioreactor-grown cell cultures.
As well, when extremely large volumes of finished product are required -- say for a monoclonal antibody that will be used to treat a chronic disease, especially one that's common -- transgenic animal and plant systems might provide the only means to achieve the necessary scale.
originally published 06/06/2002 |