November 1, 2011 (Vol. 31, No. 19)
A Strong, Predictive Expression System Is Essential to Maximize Productivity
Much has been written about volumetric productivity for therapeutic protein manufacturing processes. While focusing on titer alone ignores such factors as time, facility utilization, and product quality, a good deal of basic research continues on what makes production cells tick.
Maximizing productivity begins with a strong, predictive expression system, followed by a long development period where cells and reactor conditions are optimized. Yet what factors result in a “high producer” or “low producer”?
Expression profiling using CHO-specific microarrays has failed to uncover the answers, which must lie with varying contributions from transcription and translation rates, protein folding, and assembly. “But which combination leads higher production has been very difficult to unravel. We’re no closer to answers than we were a few years ago,” admits Mark Melville, Ph.D., director at Percivia.
He is also skeptical about cell-line engineering efforts to date, which he says have not “done a lot to push the envelope for protein expression.”
Yet basic research on expression strategies continues. Process developers purposely mimic codons in genes occurring naturally in the host system without direct data to support that choice, according to Mark Welch, Ph.D., director of gene design at DNA2.0.
This is known as codon bias: an organism’s preference for one synonymous codon over another one. Codon utilization is directly related to protein expression—not all equivalent sequences generate proteins equally. But codons purposely “biased” to accommodate the host organism do not always work best.
DNA2.0 has devised a computational platform that “surveys” the gene of interest for codon variations that produce the highest protein expression—between 10 and 100 times as much as competing services, the company claims.
“There are many ways to search for gene variants,” Dr. Welch continues. “Ours is more systematic, providing more information that covers a wider range of variables, and we can do it with fewer clones.” That means considerably fewer productivity assays.
Clone Selection
Once the gene is inside, it’s time to select the cells that express the highest protein levels at the highest quality. Clone selection resembles high-throughput screening of small molecule drugs in its reliance on the law of large numbers.
Mark Melville points to tools like Clonepix systems from Genetix, which uses fluorescence imaging to identify colonies growing in a semi-solid matrix within a Petri dish. The technique begins with ultra-high dilution that essentially separates cells individually, and ends with a liquid handler sucking up the colonies and transferring them to microtiter plates.
Melville also likes flow cytometry, another technique that relies heavily on robotic transfer of single cells to plates. Due to the low viability of isolated cells, process developers often employ tens of thousands of wells, and the technique is expensive in its acquisition and ongoing operation.
“But there have been some recent advances that incorporate omics approaches through system biology, which I expect will yield fruit within the next few years.”
Nature vs. Nurture
Despite strides in productivity, biomanufacturers remain committed to even higher titers. “I don’t think we’ve reached the ceiling yet,” says Hitto Kaufmann, Ph.D., vp of process science at Boehringer-Ingelheim.
Before process development even begins, companies need to gain a better understanding of how molecular attributes affect expression. Boehringer-Ingelheim is developing sophisticated analytics to screen for “expressability” within a bioreactor. Miniaturized (scale-down) systems are the key, Dr. Kaufmann says.
The question of nature or nurture—innate cellular capabilities versus growth conditions—always arises in these discussions.
Many cell culture experts believe that media and feed strategies have been most responsible for the run-up in volumetric productivity, and Dr. Kaufmann believes that significant, further improvements are coming. Improvement depends on techniques for rapidly screening these conditions and tailoring them to specific cell lines.
Melville believes the relative contributions of nature and nurture are equal. “You can’t separate cell line from process; they are part of the same continuum. You design the process to fit the cell line, but you also select the cell line to fit the process.”
On the “nature” side, targeted integration of genes into the CHO genome has become the new “City of Gold.” The driver here is not volumetric productivity, but faster development. “CHO cells are prone to genetic instability,” Dr. Kaufmann says. “To have designed loci within the genome that ensure stable, high expression will be a key area of focus for biomanufacturers. If successful, this approach may allow developers to skip screening for clones and media.”
Melville disagrees. “I don’t think I’ve seen any site-directed integration that has yielded higher expressions than a good cell-line screening program.” He describes a “good” screen as a reliable model that assures what occurs during clone selection will be relevant throughout process development and beyond.
“Stable integration sites and site integration have come a long way in providing a reliable, minimal expression level, but they’ve never been able to achieve ten grams per liter or higher, which we’re seeing in other systems nowadays.”
Process conditions can affect a protein product’s quality and help fine-tune its in vivo behavior—characteristics that do not translate directly into “grams per liter,” but are just as significant.
Post-translational modifications (PTMs) can affect any number of properties. Developers of therapeutic proteins have only begun to leverage knowledge of the more than 100 PTMs, which in some cases are fine-tunable by merely altering process conditions like pH or temperature.
We know that fucosylation affects the activity of antitumor monoclonal antibodies—lower levels of this sugar increase therapeutic efficacy. Similarly, higher sialylation results in a longer circulating half-life.
“The more we understand relationships between process conditions, glycosylation, and physiologic function, the more companies will exploit these strategies,” Dr. Kaufmann explains. “A few of these functional relationships are known, but many more remain mysterious, at least to the degree that we can be certain of in vivo relevance. That will change.”
In the future, companies may develop several CHO cell lines, each producing unique glycosylation patterns that tune in desired characteristics. Manufacturers of biosimilars may be able to exploit PTMs to differentiate their products as well.
The “food chain” for deploying such technology is typical for biotech. Early work and proof of concept in cells typically occurs at universities. Smaller biotechnology firms will license and optimize the relevant technologies.
If successful, they will be acquired by a much larger firm. For example GlyCart was formed in 2000 as a spinoff of the Swiss Federal Institute of Technology, Zurich. Roche acquired the company in 2005 for, among other things, GlycoMab®, a technique for directing and optimizing glycosylation in antibodies.
Alternatives to Cell Culture
Of major alternatives to expression in mammalian cells, yeast, particularly P. pastoris, are attractive for their long history of large-scale processing for therapeutic proteins and other products at scales of up to hundreds of thousands of liters.
Alder Biopharmaceuticals makes a good case for P. pastoris as a production host. When the company began operations in 2004, the founders avoided production systems that lacked an FDA track record.
“We didn’t want to face novel regulatory issues,” notes John Latham, Ph.D., CSO. Instead they set out to “change the timing, cost structure, and scale of manufacturing intact, glycosylated, fully humanized antibodies for chronic indications.”
Dr. Latham would not directly address questions about the kinds of titers Alder achieves in P. pastoris fermentations, saying only that they were “somewhat lower” than the 3–6 gram/L levels achieved with many mAb processes in CHO cells today. Shorter cycle times make up for lower volumetric productivity, however. “If you think in terms of grams of productivity per 10,000 liters per month, our first-generation process is about as good for mAbs as cell culture.”
Companies specializing in CHO culture rarely introduce the time factor when discussing titers since cell cultures take several weeks. When comparing CHO to yeast, perhaps the term “productivity cycle” makes more sense than “volumetric productivity.”
Yeast cultures have one drawback: very large harvest volumes. But for antibodies the disadvantage is minimized since initial capture occurs in flow-through mode, and recently developed capture resins allow very high flow velocities. Once the product is captured, the process resembles cell culture purification. “Except we don’t have to deal with viral inactivation or removal validation, which eats into yield due to aggregate formation. So downstream processing is actually simplified,” Dr. Latham says.
Short cycle times have the advantage of improving plant utilization. Upstream, where manufacturers plan a 10–20% failure rate, recovery and readiness to process the next batch is more rapid with yeast, and since fermentations are of short duration, downstream equipment never sits idle for very long.
Dr. Latham described Alder’s development-stage molecules as being small molecule-like in their activity. “They are designed to block pharmacologic activity of a ligand or receptor, and do not work through cytotoxicity or complement activation.” The molecules also lack N-glycosylation in the Fc region. “We engineered that out.”
Lack of glycosylation in critical regions eliminates safety issues and reduces heterogeneity. Mannose, which is abundant in yeast glycosylation, is viewed by the human system as foreign. “Eliminating mannose reduces the complexity of what lies ahead, particularly with regulators.”
By contrast GTC Biotherapeutics lets its expression system—rabbits and goats—carry out their native glycosylation. GTC was a pioneer in transgenics during the 1990s. Like most such companies it eventually ran out of money and was purchased by the French plasma products manufacturer LFB Group. The company has one marketed product, Atryn (recombinant human antithrombin), which is used to prevent clotting during surgery and as a result of several conditions. Previously, all antithrombin was isolated from plasma.
Processing throughput for transgenics has generally been good relative to conventional cell culture and fermentation, so cost of goods was not a factor in the demise of transgenics firms. The big hurdle was regulations.
“One of the concerns with recombinant proteins manufactured by transgenics was potential differences in glycosylation,” says Harry Meade, Ph.D., senior vp of research. “But analysis shows the sugars to be the same, and they are not immunogenic.”
That’s not to say that “volumetric productivity” never mattered. Expressing therapeutic proteins in field-grown corn is inexpensive; bringing plants indoors to prevent cross-pollination with nontransgenic plants raises serious cost issues.
Milk-expressed proteins provide a level of safety and economics. Over the years GTC has reportedly produced many of the top antibody products currently marketed.
Goats can routinely produce proteins, including blockbuster mAbs, at about one-tenth of a large biotech firm’s upstream cost. The two methods are on par for purification costs. “This means that we can compete with very large cell-culture processes no matter what the production scale,” Dr. Meade adds.