Protein purification rules

September 1, 2007
Susan Aldridge
Pharmaceutical Technology Europe
Volume 19, Issue 9

Getting from a cell culture to a purified biotech product is a demanding exercise involving many operations. Increasing productivity in the upstream part of biotech production is placing new demands on the purification process, which may lead to adopting new technologies.

Large-scale protein production is not easy. You may have performed extensive work on host cell development, media selection and scale-up to develop a manufacturing system that can produce, say, as much as 20 g/L of high-value product to service clinical trial supply or even commercial demand. That is where the challenge — or fun, depending on your perspective — really begins. How do you turn thousands of litres of messy cell culture brew into vials of fluffy white protein powder, while under pressure to cut costs, timelines and meet regulatory demands on purity?

While biotech clinical trial results often hit the headlines, the work that goes into purifying these drugs (so they can get into the clinic) is too often undervalued and underestimated. Yet purification can account for 30–50% of total manufacturing costs — for this reason alone, it deserves more attention.

"If you are in purification, the major new challenge is matching the production capacity to the potential increase in output from the fermenters," says Tony Orchard of Pall, a major supplier of filtration and separation products to the biopharmaceutical industry. "The other challenges of process validation, meeting timelines for new products and plant flexibility at reasonable cost are just as big."

The nature of the protein and its production system tend to dictate purification requirements. Put simply, recombinant proteins tend to be produced in microbial hosts, monoclonal antibodies in mammalian cell culture. In the former, you may need to worry about separating the product from cell debris and endotoxin contaminant, in the latter about inactivating viruses that may have the potential to infect patients.

John Liddell, purification R&D manager at UK contract biologics manufacturer Avecia explains that the initial steps in purification from a microbial host depend upon whether or not the product remains within the cell, or is secreted, when fermentation is complete. "The aim is to have the target protein in solution for the start of chromatographic purification," he says. Therefore, intracellular protein products will need to be released by some form of cell disruption, such as high-pressure homogenization. Solid–liquid separation by either centrifugation or microfiltration then removes cell debris.

The next stage is multistep (two to five steps) chromatography using orthogonal chemistries. Process development involves making each chromatographic step compatible with the previous one so product can be seamlessly loaded from one column to the next. According to a review conducted by Avecia, the most common first chromatography step is ion exchange, accounting for 75% of processes, with the second step most often being hydrophobic interaction chromatography, accounting for 60% of processes.

Although ion exchange is very useful as the first step of purification chromatography because of its high capacity, which soon reduces high volumes, Liddell argues that mixed mode media are starting to come into their own. However, mixed mode can present more process development challenges because it has more adjustable parameters to consider. There are many other new chromatographic media coming onto the market — monolithic media, beaded media with different chemistries, for instance, and media with larger pore sizes for purifying pegylated proteins (products linked to a polymer chain, which give a longer lifetime in the body).


He also notes an increasing trend towards high-throughput technologies in process development; for example, using robotic handlers with miniature columns or 96-well plates with mass spectrometry. Avecia has been evaluating the application of high-throughput approaches in overcoming some of the purification bottlenecks.

Liddell also notes a trend towards the increased use of modelling and data mining techniques in purification. Proteins are notoriously fickle and it is still impossible to predict behaviour and characteristics from sequence. But misfolding, aggregation and proteolytic instability in a protein molecule can cause real headaches in purification and it is better to be forewarned, if at all possible. "Developments in structural biology will allow more and more prediction of purification relevant information from sequence," Liddell predicts. "More initial purification development is likely to be performed through modelling prior to starting experimental studies."

There is also increasing interest in data mining, for individual companies and contract manufacturers, he adds. Avecia has information on more than 35 biologic production processes, which can be used to inform future processes.

Novel approaches

There are alternatives to chromatography for large-scale purification of proteins from both microbial and mammalian host cells, such as precipitation and crystallization. Recent reports from some companies suggest there is some cautious experimentation being conducted with these approaches. However, there is an understandable reluctance to abandon tried and tested chromatographic cascades that are approved by the regulatory bodies unless driven to by the demand of increased process volumes.

However, such radical change may prove necessary. Most of the growth in biotech production is currently focused on monoclonal antibodies, with approximately 160 of these protein molecules in development. A monoclonal is typically captured on a protein A column in initial purification. This step, along with viral inactivation, accounts for a major part of the cost of purification. With antibody titres increasing, there is also concern regarding whether protein A is beginning to approach its capacity. Therefore, for cost and technical reasons, finding replacements for protein A is becoming something of a pre-occupation among manufacturers. There is no obvious solution, but some alternatives were summarized by Duncan Low, scientific director of process development at Amgen, at a recent meeting. These include bacterial immunoglobulin G binding protein, lectins, immunoaffinity columns, single domain antibodies (highly selective and the closest to protein A) and synthetic ligands. Less complex ligands are cheaper, but tend to be less selective towards the target.

"Assuming that monoclonals continue to dominate and grow in demand, there will be some trade-off between the resistance to change from proven production platforms and the need for more drastic changes driven by volumes, costs, opportunity and competition," predicts Orchard.