Quality or Yield? Or Both?

Compromises between product quality and manufacturing productivity are common in consumer industries. “You get what you pay for” works, for the most part, in how the invisible hand of commerce sorts out the fates of high-priced, high-quality goods and those with built-in obsolescence.

No such luxury exists for biopharmaceuticals, where the price of shoddy manufacture is poor patient outcomes.

A hundred out of a hundred producers of therapeutic proteins will make this point. What is less well appreciated in today’s competitive, risk-based manufacturing environment is the reality that a product’s characteristics are described in terms of ranges, or three-dimensional “spaces,” rather than specific targets. In other words wiggle room exists.

Compromises

The manufacture of therapeutic proteins through mammalian cell culture very frequently entails compromises. We know, based on simple mass balance, that resources in media and feed are limited as are a cell’s resource allocation capabilities. One often observed trade-off exists between yield or productivity and cell growth, as noted by Prof. Nathan Lewis at the University of California, San Diego.

Biocompare’s Cell Culture Media Search Tool
Find, compare and review media
from different suppliers Search

Last year Lewis reported on a genome-scale model of Chinese hamster ovary (CHO) cells that identifies protein production pathways and predicts productivity of CHO cultures under two conditions, lower culture temperature and addition of sodium butyrate, which have been reported to improve productivity in some systems, and not in others. “Sodium butyrate opens up chromatin to allow activation of suppressed genes,” he explains. The model predicted that the treatments would in fact improve yields but only slightly, and at the expense of cell growth.

The productivity-cell growth tradeoff is inevitable, he believes, because carbon sources (and presumably cellular energy and operations) are expended on cellular reproduction.

Better, Lewis says, to target mechanisms that facilitate protein synthesis, folding, and secretion to increase yield and, in many instances, improve quality as well.

Where the tradeoff between yield and cell growth is understandable by virtue of simple mass balance, the compromise between quality and productivity is more mysterious. Common wisdom currently holds that genome editing has greater potential for improving yields than improving quality.

Prof. Lewis believes the issue is somewhat more complex. He notes that attempting to push high levels of product through a cell’s manufacturing apparatus might overwhelm glycosyltransferases, the enzymes responsible for attaching desirable glycans to a glycoprotein. “This could alter critical quality attributes in unpredictable ways.”

This should not be a problem, however, unless the up-titering saturates glycosylation pathways. For strategies based solely on increasing cell density this should not be the case as each new cell brings, under normal circumstances, the apparatus to produce and glycosylate. As long as these are balanced to begin with no deficit should result from higher cell density.

“But with higher cell density comes a greater concentration of host cell proteins (HCPs), and you can’t tell beforehand how these might affect protein quality,” Lewis says. At higher concentrations of both potentially degrading or inhibitory HCPs and their substrates, interactions that negatively affect quality (or yield) might predominate in ways they do not at relatively low concentrations of both. “We have unpublished data suggesting this,” he adds.

The Lewis group has been developing further models to guide efforts to engineer glycosylation, a critical quality attribute for monoclonal antibodies and other glycosylated proteins. The team has devised an algorithm that predicts how researchers can modify CHO cells to obtain desired glycosylation patterns during manufacture of innovator biologics or biosimilars, by genetically modifying the cells or changing the media composition. These efforts can help in coaxing cells to produce proteins with more favorable fucosylation, sialylation, and glycan branching.

Have your cake and eat it

In early 2017, ATUM (formerly DNA2.0) entered a cross-licensing agreement with Horizon Discovery to combine that company’s CHO SOURCE cell lines with ATUM’s vector technology for introducing protein-associated genes into those cells. Thus both companies’ customers have the means to express recombinant proteins in a manufacturing-worthy expression system.

Because the combined technologies address the relevant topics of this article, they potentially obviate the conundrum between quality and productivity.

CRISPR Cas9 gene editing is now all the rage but, according to Claes Gustafsson, Ph.D., CCO of ATUM, the technology is much better suited to deleting genes than inserting them. While gene deletion can address some quality issues, for example the expression of undesirable host cell proteins or an inhibitor of proper folding or excretion, it cannot specifically add characteristics with the reliability required for the production of therapeutic proteins.

ATUM instead relies on transposons, or jumping genes, discovered by Nobelist Barbara McClintock in corn.

Gustafsson discovered diamonds-in-the-rough versions of transposons by searching GenBank, the NIH gene sequence repository. Of hundreds of transposons examined two were capable of reliable gene insertion and were not covered by industrial patents. After optimization these constructs became the foundation of ATUM’s Leap-In product, which, according to Gustafsson, “allows creation of pools of high-expressing cell lines in about two weeks.”

If that were all the ATUM-Horizon collaboration achieved it would be noteworthy. The ability to insert and delete genes in a high-expressing host has even greater implications for the quality/yield question. “Now all the cell’s machinery is open to engineering,” Gustafsson says. “We can for example knock in and knock out glycosylation or secretion machinery through orthogonal approaches. Our mAb engineering platform allows you to have your cake and eat it too.”

Through machine learning and synthetic biology ATUM engineers antibodies in multiple functional dimensions in parallel, for example glycosylation, target binding, titer, aggregation, and thermal stability, which compose the broadly defined “quality machinery” of cellular expression systems.

“We’ve seen examples of cells producing way too much protein for its glycosylation machinery to handle,” Gustafsson explains. “In those instances we can knock in additional glycosylation machinery.”

This approach demolishes the existing paradigm, where cells are sequentially screened, beginning with affinity, for various orthogonal qualities that barely overlap. “A zillion different antibody sequences can bind strongly, but most are poorly expressed. Instead of randomly screening for those qualities you can engineer them into the antibody in parallel.”

Pulling all stops

It will take a few more years, but combining gene editing with media and feed strategies will probably some day eliminate the quality-yield question. In this paradigm clone selection should not be overlooked.

Selexis specializes in cell-line development through its SUREtechnology Platform™, which improves clone selection for whatever specific qualities a manufacturer desires. Company president Igor Fisch bristles at the notion that quality must ever be sacrificed for yield. His work, he says, “has been predicated on improving yield while maintaining quality; each cell clone is chosen based on product activity and yield. If the product activity does not meet specs determined by preclinical work the clone is not chosen,” adding that doing otherwise would be “too risky.”

Biomanufacturers first lean toward quality. However, they must be able to manufacture that quality at a level that is economically feasible.

A Moot Question

Lonza continues to improve its GS Gene Expression System®, which first appeared 30 years ago. Upgrades led to the GS Xceed® System, which generates high titers in a chemically defined medium that does not require methionine sulfoximine (MSX) selection. GS Xceed® typically achieves titers of six grams per liter, with 10 grams per liter reported.

The question of quality-yield tradeoff often arises with such high-producing systems. Andy Racher, associate director for future technologies at Lonza, explains why: “There are occasions where you have to dial back productivity because the product lacks certain quality characteristics. This typically occurs for more complex proteins.”

Manufacturers of standard antibodies amenable to platform manufacturing can usually go for the highest titer but for fusion proteins, or proteins with complex glycosylations (e.g. a combination of N- and O-glycosylations) manufacturers must optimize the process or ramp it down by changing culture temperature, harvesting earlier, or slowing down cell growth.

“Critical quality attributes could moreover be different for non-standard proteins,” Racher explains. For example, high sialylation is often desirable in fusion proteins but not really needed for antibodies. “Under certain conditions these products are more prone to fragmentation and since they are less ‘natural’ they may not fold properly.”

Racher believes that the potential to improve titer through media and feed strategies may still exist, but in the context of quality-vs.-yield that question is almost moot. “There could indeed be a “magic” ingredient that might enhance desirable characteristics and inhibit undesirable ones, while improving productivity, but we tend to view media and feed as an inherent part of the process. You can’t really separate the two because the process controls how cells handle nutrients and remove metabolites.”

Image: Dreamstime Images