January 1, 2013 (Vol. 33, No. 1)
The manufacture of protein therapeutics is only as efficient as the expression system. Despite the emergence of several platform approaches (e.g., CHO for monoclonal antibodies), drug developers are constantly tweaking cells and microorganisms to optimize productivity and product quality.
One focus of CHI’s upcoming “Peptalk” will be expression, and this article features several scheduled speakers on the topic.
Kevin Sunley, Ph.D., senior scientist at XBiotech, discusses some remarkable observations on the effect of GE Healthcare’s Cytopore™ microcarriers on CHO cell productivity. Cytopores are porous, hollow, positively charged (due to DEAE functionalization) spheres into which anchorage-dependent cells enter and grow. The idea is to provide very high surface areas for attachment, an environment that protects cells against shear forces, and porosity to allow nutrients to enter and waste products to exit. Dr. Sunley refers to the microcarriers as “mesh snowballs.”
He says the microcarriers work as advertised: Cell-specific productivity rose in proportion to the increase in cell density within the carriers, providing a multifold increase in protein yield compared with stirred-tank suspension culture. The surprise is that the dramatic improvement was not due to cell attachment within the carriers. Upon examination the cells “look like suspension cells, so the yield improvement must be due to some other biological function,” Dr. Sunley says.
He theorizes that the 200-micron spheres become a sort of spa and resort for the cells, which enter and form very dense colonies while remaining in suspension mode. Cell densities reach 4×108/cc, slightly above the theoretical density. When carriers become overgrown, cells leave and populate empty carriers—like aliens seeking out new worlds—which would not be possible with attached cells.
“The key idea from this project is not just that cells are inside the carriers and protected, but that they are in an environment that is different from the outside environment. It’s a very favorable local cell environment independent of the bioreactor.”
Large stirred-tank reactors, he adds, are far from homogeneous environments. “If you measure pH, for example, you’re obtaining a reading at the probe, which may not be what the cells are seeing at all. Here we’re giving the cell a unique environment that is protected, high-density, and nutrient rich. And since the carriers are open, the cells almost have their own perfusion system.”
These findings led Dr. Sunley’s group to consider cell “happiness” as a primary consideration in the design of cell culture. He believes that Wave bioreactor bags, another GE product, provide greater opportunity for creating a sustaining, heterogeneous environment within the culture than much larger stirred stainless steel tanks.
Confocal micrograph of a Cytopore™ microcarrier populated with recombinant CHO cells. Microenvironments within the bead are illustrated within the higher magnification optical section. [University of Manitoba]
Glycosylation Twist
Darius Ghaderi, Ph.D., senior research scientist at Sialix, points out challenges of nonhuman sialylation in biotherapeutic glycoproteins. We know that glycosylation is, in many instances, critical for a protein’s therapeutic efficacy, immunogenicity, overall quality, and even manufacturing yield. The success of mammalian cell culture in the production of monoclonal antibodies is in part due to the cells’ almost human-like glycosylation. Nevertheless, subtle differences in the glycan structure, particularly at the most exposed terminal positions, can have profound effects.
Dr. Ghaderi is most concerned with two anomalies common in nonhuman expression systems: the terminal Galα1-3Gal (alpha-Gal) and N-glycolylneuraminic acid (Neu5Gc) modifications. Humans express antibodies against both glycan structures, which carries the risk of immunogenicity.
Interestingly, cultured human cells incorporate Neu5Gc but not alpha-Gal. “Humans lost the gene for biosynthesis of alpha-Gal 20–30 million years ago, but the loss of Neu5Gc biosynthesis in humans was more recent, about 2–3 million years ago,” Dr. Ghaderi says. Other mammals retained the Cmah gene, which is responsible for the presence of Neu5Gc, to this day.
Loss of the gene coding for alpha-Gal biosynthesis explains why humans do not express or use alpha-Gal at all. Contrary, exogenous Neu5Gc can be incorporated into human tissues, even though humans produce antibodies against this foreign sialic acid.
Some implications of this work involve the human dietary intake of Neu5Gc and exposure to tick bites. A recent study suggested that humans may incorporate Neu5Gc from eating the meat of animals that comprise this non-human immunogenic sialic acid.
The resulting inflammatory response to incorporated Neu5Gc might be responsible for downstream inflammatory processes that lead to chronic diseases. In addition, receiving Neu5Gc-containing biotherapeutic glycoproteins might cause serious adverse reactions in patients. FDA-approved Erbitux, for example, contains both alpha-Gal and Neu5Gc “in relatively high concentrations,” according to Dr. Ghaderi. Moreover, ticks were recently found to incorporate Neu5Gc into endogenous salivary glycoproteins after feeding on mammals. A subsequent bite to a human might therefore transfer such Neu5Gc-containing saliva and elicit an immune response against Neu5Gc. Tick bites have already been correlated to meat allergies and anti-Gal mediated immunogenic responses to Erbitux.
Nonhuman glycan structures in available glycoprotein expression systems showing the major differences between human glycosylation and nonhuman glycosylation. [Sialix]
Proteins from Prokaryotes
James A. Ernst, Ph.D., scientist, departments of protein chemistry and early discovery biochemistry at Genentech, describes a membrane protein expression in bacteria. Dr. Ernst’s technique involves “novel transcription control elements that allow the successful application of translational regulation to the expression of integral membrane proteins in E. coli.”
The basic secretory machinery for eukaryotes (e.g., yeast, mammalian cells) and prokaryotes (bacteria) is similar, with some subunits of the machinery transferable from yeast to E. coli. Yet scientists have been puzzled by the difficulty of expressing eukaryotic membrane proteins in a bacterial host.
“One problem is cytotoxicity as the desired proteins are expressed,” Dr. Ernst says. “This results, in part, in how proteins interact with the secretory system. Bacterial translation is much faster than in yeast, but the coupling between translation and membrane insertion is not as reliable, which gums up the translation mechanism.” He hypothesizes that cytotoxicity results from a disconnect between normal bacterial translation and membrane protein insertion, with protein aggregation as the main consequence.
The literature contains examples of attempts—some successful, some not—to express membrane proteins in mammalian or other eukaryotic expression systems. Still, bacterial systems have the advantages of accessibility, low cost, high expression levels, and shorter culture times. “We can use bacterial host for many other proteins. Why not for these?” Dr. Ernst asks.
His solution was to create a “tighter” promoter that more effectively coupled translation and membrane insertion. Most bacterial promoters are either positively or negatively regulated. Positive regulation involves the promoter binding a substrate, and the transcription factor moving onto the DNA—similarly to how many eukaryotic systems operate. In contrast, most bacterial processes are negatively regulated by elements that repress transcription.
“We took those two systems and combined them, creating a much more tightly regulated system that avoids cytotoxicity. The approach provides a test bed for evaluating other theories about what was affecting membrane protein expression in bacteria, including regulated translation initiation.”
Practically speaking, Dr. Ernst’s method allows control of the secretory process for many drug targets of interest to Genentech, for example CD20, the target of the breast cancer antibody Rituxan.
“Our ability to make and evaluate proteins experimentally is improved by having new methods by which to make the protein in either bacterial or mammalian cells. Plus it improves our ability to discover new antigens and antibodies.” Dr. Ernst worked on this project with Genentech collaborators Dan Yansura and Hok Seon Kim.
Parallel Processing
The Pfenex Pseudomonas fluorescens-based expression platform is well-known. Pfenex produces difficult-to-express proteins by applying “parallel processing” to expression strain development. According to the company, this replaces the “traditional, linear, and iterative approach to strain development with a high-throughput, parallel processing model that allows construction and testing of 1,000s of unique expression strains combining novel gene expression strategies and host cell phenotypes” in less than five weeks.
Titers may be as high as 20 g/L. Because development times are short, Pfenex also supports early-stage development protein demands for characterization studies.
In October, Pfenex awarded Althea Technologies a contract for the GMP manufacturing of circumsporozoite protein, an important component of malaria vaccine. Althea will deploy a Pfenex Expression Technology™-based production process that was developed at Pfenex, which is also developing processes for additional malaria antigens.
Just a week earlier, Pfenex received a National Institute of Allergy and Infectious Diseases contract to develop an alternative delivery method for Pfenex’ recombinant protective antigen (rPA)-based anthrax vaccine. The initial award of $2.18 million could be supplemented up to $22.9 million if the parties exercise all options. This grant supports development of a dry formulation of a Pfenex-rPA anthrax vaccine with very long shelf life and suitability for needle-free delivery.
Unraveling Codon Bias
Bioinformatics has improved the ability to manipulate gene sequences and explore protein sequence variation through control over codon bias and mRNA structure, says Mark Welch, Ph.D., director of gene design at DNA2.0.
Genes provide researchers with copious information that includes protein coding data. Since individual amino acids are generated by up to six different codons, more possibilities exist for a typical protein than the number of atoms in the universe. Optimizing protein expression through systematic manipulation of codons is therefore impossible.
Yet discovering an “optimal” codon, given time and financial considerations, is highly desirable. Codons affect protein expression levels through several mechanisms, including RNA longevity. Effects may arise at the translation level as well, as certain codons translate faster or may be more familiar to cells.
“According to conventional wisdom, the codons the cell uses most often, that the cell devotes the most machinery to, are typically the fastest,” Dr. Welch explains. But instead of assuming that those codons are the best, DNA2.0 explores codon usage systematically, with the knowledge that optimization has the potential to improve expression levels by two orders of magnitude. “We see correlations between codon usage frequencies and expression that disagree with what traditional codon optimization assumes—that the most frequent codons are the fastest.”
Due to the sheer number of possibilities, Dr. Welch is reluctant to use the term “optimized” to describe gene sequences he generates. “We don’t sample the entire space. We’re not even close.” Rather, his strategy creates a codon sequence that is “optimal within the rules we’ve identified as significantly affecting protein expression.”
Benefits of codon-tweaking are not limited to productivity. Dr. Welch points out that for some manufacturing processes, like those for making monoclonal antibodies, productivity is not an issue. In those situations DNA2.0’s approach can improve secretion and folding. “That’s a high-level definition of optimization: not just total protein, but total active protein. Our method covers whatever the customer’s definition of optimal may be.”
