Next-Gen Expression Systems

Most of the biopharmaceuticals that have reached the market in the United States and the European Union are produced in just a few expression systems. A recent report in the September 2010 issue of Nature Biotechnology indicated that 32 of the 58 biopharmaceuticals approved between 2006–2010 were produced in mammalian expression systems (mostly Chinese hamster ovary [CHO] cells), and 17 were produced in Escherichia coli (E. coli). The remainder were produced in yeast (4 in Saccharomyces cerevisiae, 1 in Pichia pastoris), transgenic animals, insect cells, plants, or by direct synthesis (1).

Choosing the best expression system depends on the complexity of the product, the batch size, and cost constraints, but each system comes with tradeoffs. E. coli was the first system used for commercial production, with the marketing of human recombinant insulin in 1982 (2). The bacterium is easy to culture, proliferates rapidly, and can be grown in inexpensive, defined media. In the nearly 30 years since, manufacturers have continually improved the system, optimizing culture conditions, scaling production volumes, and boosting protein yields. E. coli expression systems are limited, however, in the types of proteins they can produce, and are best used for simple proteins with few posttranslational modifications. E. coli does not produce glycosylated proteins, for example, nor can it correctly produce large proteins with complicated folding patterns, or multiple disulfide bonds. E. coli does not secrete proteins, requiring the manufacturer to take extra purification steps, including cell lysis, centrifugation, multiple filtrations, and denaturing and subsequent refolding of the protein. Gram-negative bacteria such as E. coli also contain endotoxins, which must be removed.

IMAGE: MEDICALRF.COM/GETTYIMAGES

Mammalian cells have become the system of choice for antibody production, which requires most of the posttranslational modifications that can't be achieved in E. coli. Mammalian cells—CHO cells being the most widely used— are best equipped to secrete glycosylated, multi-subunit proteins that will fold and assemble correctly. They are amenable to genetic manipulation, and lines exist that have been optimized for ease of transfection, increased protein expression, and high-density culture. The major drawback to mammalian cells is that compared with microorganisms, they are more difficult to culture, require expensive culture media and a low-shear environment, and they often produce lower yields. There are also biosafety concerns around the transmission of viruses that can be pathogenic in humans.

Some more recent entrants on the market strive to address some of these drawbacks. Most notably, PerC.6, a transformed human cell line jointly developed by Crucell and DSM for use as a manufacturing-scale expression system, increased expression yields so dramatically, that it shifted manufacturers' concerns from producing adequate protein yields to having adequate capacity to efficiently purify the increased titers. PerC.6 can be grown in suspension cultures at relatively high densities in serum-free medium, and according to company literature, can achieve yields of up to 8 g/L in standard fed-batch cultures, and 20 g/L using DSM's proprietary culture system, which are levels comparable to those of microbial expression systems.

Building better bacteria

While E. coli has been a mainstay of biopharmaceutical manufacturing, other bacteria may offer certain advantages over E. coli. For instance, Corynebacterium glutamicum is a gram-positive soil bacterium that has been used for decades in the food and pharmaceutical industries to produce amino acids, most notably, monosodium glutamate. Unlike the traditional E. coli system, C. glutamicum possesses secretory pathways that can be coopted to secrete engineered proteins into the culture broth. Marketed as the Corynex expression system by Ajinomoto AminoScience, this strain secretes few endogenous proteins, but engineered proteins can be targeted to one of two secretion pathways, the secretion pathway (sec), or the twin arginine translocase pathway (tat). The sec pathway initiates secretion of unfolded proteins that subsequently fold in the broth, while tat is able to transport folded proteins, and thus is able to handle larger proteins. Expression is constituative, and no inducer system is required. Joel White, business manager of biotechnology at Ajinomoto Aminoscience, says that the growth of Corynex is robust: it has a rapid doubling time, only slightly slower than E. coli, and can be grown in inexpensive, defined media in standard bioreactors. The great advantage of the system, according to White, is that the extra steps of cell lysis, centrifiguation, and protein refolding are not necessary, thereby providing a simpler, faster, less expensive purification route. White adds that this strain broadens the spectrum of folded proteins that can be produced in a bacterial expression system.

Other companies have been engineering E. coli to overcome its limitations in producing eukaryotic proteins. New England Biolabs, for example, markets several engineered strains of E. coli that are used primarily for bench-scale protein production. One of the strains, known as SHuffle, has been engineered to allow disulfide bond formation in the cytoplasm. In wild-type E. coli, reductases in the cytoplasm inhibit disulfide bond formation by keeping cysteine residues in their reduced form. Two reductases have been deleted in the SHuffle strain, and a version of a periplasmic disulfide bond isomerase has been targeted to the cytoplasm, which helps correct misoxidized bonds and promotes proper folding. Company data show that the SHuffle strain is capable of producing biologically active truncated tissue plasminogen activator, a protein that contains nine disulfide bonds when correctly folded. According to Dr. James Samuelson, a staff scientist at New England Biolabs, the strain grows well in fermenters and produces high protein yields. While most interest in the strain has come from academic labs, other companies, including manufacturers, have also expressed interest.

Controlling glycosylation

Bacterial systems grow quickly, but can generally not produce large proteins that require extensive posttranslational modification. Mammalian systems grow more slowly, but can deal with posttranslational modifications, and yeast fall somewhere in between. Yeast grow rapidly and can be cultured inexpensively, and unlike bacteria, secrete glycosylated proteins through the Golgi apparatus. However, the pattern of glycosylation in wild-type yeast differs markedly from the mammalian pattern and can alter a drug's potency or create unwanted immunogenic responses.

Developing yeast strains that produce human-like glycosylation patterns required extensive genetic engineering. Researchers had to first suppress the endogenous yeast glycosylation systems, and then provide the necessary enzymes to produce human-like carbohydrate chains. Working in the yeast strain Pichia pastoris, a team of researchers at New-Hampshire based Glycofi and Dartmoth University were able to reproduce the N-glycosylation pattern found on human glycoproteins. This was done by deleting four yeast genes, and inserting 14 heterologous genes, producing a humanized carbohydrate chain that contained sialic acid attached to the terminus of the chain, which is critical for metabolic stability in vivo (3). Glycofi and its humanized P. pastoris platform were acquired by Merck in 2006 specifically to support Merck's biosimilar-development program.

Other companies have since stepped up to provide expression systems where glycosylation patterns can be controlled. Wild-type CHO cells produce proteins with a heterogeneous composition of carbohydrate chains, and the factors that control the extent of glycosylation and type of carbohydrate chains attached to any given protein are not well understood (4). CHO cells have been engineered through a collaboration between BioWa of New Jersey, and Lonza to produce a cell line, called Potelligent, which is a suspension-adapted CHO cell line that, lacking the enzyme fucocyl transferase, produces glycoproteins without fucose. Non-fucosylated antibodies have been found to exhibit more effective antibody-dependent cell-mediated cytotoxicity than fucosylated antibodies (5). Company data from Lonza have confirmed this finding for multiple antibodies. While glycoengineering has not proceeded to the same extent in mammalian cells as it has in yeast cells, a cell line such as Potelligent demonstrates that control of glycosylation patterns can have dramatic effects not only on protein stability and immunogenicity, but also on efficacy.

Expression systems for biobetters

Several groups have been working on ways to develop protein therapeutics in the same way that small-molecule therapeutics are developed. That is, they introduce systematic changes to the molecule, and assay how the changes affect properties, such as metabolic stability, affinity, and efficacy. However, protein biochemistry is not as flexible as small-molecule chemistry. The biological activity of a protein depends on correct three-dimensional structure, and small changes in protein stereochemistry may dramatically affect protein stability, affinity for its receptor, or efficacy. Moreover, the amino acids that make up mammalian proteins have a limited capacity for chemical modification. Reactive side groups are found on lysine and on cysteine, but there is no way to target a single amino acid for modification. Any modification targeted, for instance, to lysine, would potentially affect all the lysines in the protein, but would be more likely to produce proteins with a heterogeneous mixture of modified and unmodified sites.

The ability of protein chemists to modify the naturally occurring form of a protein has been advanced by Ambrx of San Diego. They have developed manufacturing-scale expression systems that allow insertion of nonnatural amino acids into defined positions in the protein, says Dr. Ho Cho, chief technology officer at Ambrx. The purpose of doing so is two-fold. First it allows one to introduce targeted changes to the protein, and to determine how these changes affect protein structure and function. Second, it allows manufacturers to use the nonnatural amino acid as a site that can be selectively chemically modified. The technology was originally developed in E. coli, and marketed as the ReCode expression system, which stands for Reconstituting Chemically Orthogonal Directed Engineering. It has since been expanded to eukaryotic cells, (e.g., yeast and CHO cells), and is marketed as the EuCode expression system. The EuCode expression systems are capable of producing fully glycosylated, multi-subunit proteins, such as antibodies and blood factors, that incorporate nonnatural amino acids, says Cho.

In the EuCode system, cells contain an orthogonal transfer RNA (O-tRNA) that will read through (suppress) a stop codon called amber. Ambrx has engineered tRNA synthtases that will amino acylate the O-tRNA with an Ambrx nonnatural amino acid. When the ribosomal complex encounters the amber stop codon, the amino acylated O-tRNA will insert the nonnatural amino into the elongating protein. The O-tRNA and tRNA sythetases are orthogonal, because they do not interfere with the cells' endogenous tRNAs, tRNA synthetases, or the incorporation of naturally occurring amino acids into the protein. In the absence of an Ambryx amino acid, protein synthesis is terminated at the amber codon.

The expression system is currently used to introduce a single amino acid substitution at a specific site to minimize disturbance to the rest of the protein, to preserve the protein's normal structure and function. Orthogonal tRNA synthetases have been developed for >50 nonnatural amino acids that possess unique functional properties. The nonnatural-amino-acid substitution can be used to explore structure-function relationships in fully glycosylated proteins, or as sites of attachment for effector molecules to enhance protein function, as in the case of drug-antibody conjugates or improve metabolic stability. The ReCODE system has been successfully scaled to >50,000 L fermentation with consistant product titers of >5 g/L.

California-based Sutro Biopharma has been attempting the same type of manipulation, but using a cell-free expression system. As a bench-scale tool, cell-free expression systems have been in existence for many years, but recent advances, in particular, identification of a cost-effective supply of energy to power the system, have made it possible to scale up these systems to achieve yields purported to be suitable for manufacturing. Sutro Biosciences has developed a cell-free expression system that can be scaled from 200 μl to 200 L. The system uses an extract of E. coli (KGK10) that contains the cell's protein synthetic machinery, but is free of bacterial DNA. The extract can be produced in bulk and frozen for up to a year before use. The reaction mix contains the bacterial extract, a source of energy, amino acids, and the DNA coding the protein to be produced (6). The time- consuming processes of cell transfection and clonal selection are eliminated.

The system can be used to produce large molecular weight, correctly folded proteins with multiple disulfide bonds as well as multi-subunit proteins, according to Dr. Trevor Hallam, the company's chief scientific officer. Using this system, it has been possible to produce a number of proteins that have been difficult to express in E. coli. For example, Sutro can now produce up to 1300 mg/L of a version of biologically active human recombinant granulocyte-macrophage colony stimulating factor. The in vitro rate of protein synthesis is around one-twentieth the in vivo rate, which seems to facilitate the correct folding of the proteins.

An open system allows for certain manipulations that would not be possible in vivo, notes Hallam. For instance, the stoichiometry of a multi-subunit protein can be controlled by adjusting the ratio of DNA templates that are added to the reaction. Temporal control of expression of a multi-subunit protein is also possible. When expressing antibodies, Sutro researchers have found that adding the DNA template for the light chain an hour in advance of that of the heavy chain prevents aggregation of the heavy chain protein.

The cell-free system offers the potential to do development and manufacturing in the same system. Working initially at 96-well scale, researchers would be able to rapidly screen many iterations of the protein of interest. Nonnatural amino acids can be easily introduced, since there is no barrier to uptake, with the amino-acylated tRNA required for insertion added to the reaction mix as a reagent. Developers can then vary the site of insertion of the amino acid, and explore structure–activity relationships just as they would when developing small-molecule therapeutics. Once the optimal structure is identified, scale-up to commercial levels would be straight-forward, says Hallam. By using the non-natural amino acid as a point of attachment for an antibody–drug conjugate, drug developers could control the number of drug conjugates per antibody to produce a single species of high quality.

The current dominance of just two expression systems, E. coli and CHO cells, for biopharmaceutical manufacturing is driven by the fact that both manufacturers and regulatory agencies are familiar with those systems. The increasing market in biosimilars should move manufacturers to explore and adopt alternative expression systems, if there are cost-savings to be realized by doing so. Additionally, biologics developers are becoming more sophisticated in the types of molecules they wish to design. The market has moved beyond simple protein therapeutics to encompass antibodies, antibody fragments, and drug–antibody conjugates, which will drive adoption of systems that allow more control over the finished product.

References

1. G. Walsh, Nat. Biotechnol. 29(9), 917–924 (2010).

2. FDA, "Celebrating a Milestone: FDA's Approval of First Genetically-Engineered Product."

http://www.fda.gov/AboutFDA/WhatWeDo/History/ProductRegulation/SelectionsFromFDLIUpdateSeriesonFDAHistory/ucm081964.htm, accessed May 2011

3. S.R. Hamilton et al., Science 313(5792), 1441–1443 (2006).

4. P. Hossler, S.F. Khattak, and Z.J. Li, Glycobiol. 19(9), 936–949 (2009).

5. Y. Kanda et al., Glycobiol. 17(1), 104–118, (2006).

6. Zawada et al., Biotech. Bioeng. 108(7), 1570–1578, (2011).