Eluting Possibilities with Mixed-Mode Chromatography

May 2, 2007
Pharmaceutical Technology
Volume 31, Issue 5

Mixed-mode chromatography sorbents and custom ligands aim to optimize protein purification.

Changes in the biopharmaceutical industry have led to the need for new chromatographic techniques beyond traditional methods such as hydrophobic interaction chromatography, (HIC), ion exchange chromatography, and affinity chromatography. Protein expression titers and feedstock volumes have increased greatly during the past decade, and fermenter volumes of 12,000–15,000 L are not uncommon.

"Some of the technologies that were developed about 30 years ago are no longer optimal for some of these new challenges," says Sylvio Bengio, scientific communications manager for Pall Life Sciences (East Hills, NY, www.pall.com~).

Conventional techniques, including the Protein A affinity, still work well. But the cost of these methods is weighing heavily and has led to increasing interest in alternatives, especially new sorbents for mixed-mode chromatography and specially designed ligands.

Improving mixed-mode sorbents

In principle, mixed-mode chromatography is not new. However, researchers now are able to specifically design and finely "tune" the hydrophobic dimension of the mixed-mode mechanism.

Converting a dilute feedstock to a high-purity protein typically involves three steps: a capture step, an intermediate step, and a polishing step, which may involve ion exchange, but there are situations where this is no longer ideal. One example is when processing large volumes of feedstream. Typically, to obtain sufficient capacity with conventional ion exchangers, the conductivity of the feedstock must be adjusted by dilution, sometimes to as much as five times the original volume. Although this practice is feasible at the laboratory scale involving 5- or 10-L volumes, working with a 10,000-L feedstream is another challenge. "Obviously diluting three or four or five times is totally unacceptable because buffer storage volumes create problems," says Bengio. "Ideally, at the capture step, you would like to have a minimal dilution of the feedstock and still get good capacity and recovery in one step; this is exactly what new mixed-mode platforms will do in most cases."

Monoclonals currently represent more than 35% of the proteins in clinical trials, and the current platform technology is affinity chromatography on Protein A. Protein A sorbents are very selective for capturing antibodies from complex feedstreams with high purity in just one step. One of the biggest problems with this ligand, however, is its cost. Some antibody processes may require several hundred kilograms of pure mono clonal for different treatments. "There is strong pressure on the industry to cut purification cost," says Bengio. "Chromatography and especially Protein A sorbents are one of the major contributors to purification cost that the industry wants to cut."

Mixed-mode chromatography also has been used in place of or in combination with HIC processes (i.e., at the initial HIC step, in which case the methods would be orthogonal). Although conventional HIC is well-known, broadly used, and effective at various stages in purification, it requires the addition of high concentrations of lyotropic salt, typically ammonium sulfate, to the feedstock to promote protein binding to conventional HIC resin. Concentrations can range from 1 to 3 M. At the laboratory or small scale, this is usually not a problem. "But in industry where there are thousands of liters of feedstream, adding massive quantities of salt is a nightmare because salt is expensive and creates hardware problems. Furthermore you have to recycle that salt for environmental protection purposes, and this increases the overall cost" says Bengio.

Russell Jones, chromatography marketing manager at Pall Life Sciences, notes that a company using conventional HIC sorbents, is likely to use several tons of lyotropic salt per year. "A process column containing hundreds of liters of conventional HIC sorbent will require the use of significant volumes of buffers containing 1–3 M lyotropic salt, especially over the course of an annual campaign of 30 cycles or more. This will not only have an impact on process cost but also on what you have to do afterward to dispose of the salt in an environmentally friendly way."

Purification methods: old and new

Not only is Protein A expensive, but it leaches and must be removed. Furthermore, the Protein A resin is not stable to 1 M NaOH, which is frequently used for cleaning. (Although GE Healthcare recently introduced "MabSelect SuRe," which is stable in 0.5 M NaOH.) Also, the method requires elution at low pH, which tends to cause aggregation and damage the antibody.

BioSepra (now Pall Life Sciences) introduced "MEP HyperCel" as the first mixed-mode resin (originally developed to replace protein A for antibody capturing). The ligand, MEP (4-mercaptoethylpyridine) contains a heterocyclic ring and a thioether linkageknown to have an affinity for antibodies. MEP HyperCel allows antibody binding at neutral pH directly out of the feedstock and, unlike protein A, binds all species and isotypes, including IgM. Elution can be carried out at pH 4 or higher.

Andrew Lees, scientific director at Fina Biosolutions LLC (Rockville, MD, www.finabio.com) explains that elution from MEP HyperCel is conducted by reducing its pH. As the pH decreases, the pyridine ring picks up a positive charge, so the sorbent goes from being hydrophobic to one that has a positive charge (similar to an anion exchanger).

Decreasing the pH also causes the protein to become more positively charged, so it is repelled. Proteins that are more basic (i.e., have a net positive charge), including most antibodies, are eluted first and, as the pH decreases, more acidic proteins are eluted. In addition, the column also retains its hydrophobic characteristics. Therefore, the first materials that will elute are those that are more basic and more hydrophilic; materials that tend to elute last are more acidic and more hydrophobic (see Figure 1). Because antibody aggregates tend to be more hydrophobic compared with the monomer, they will be more tightly bound to the sorbent and so will elute at a lower pH than the monomer.

Figure 1

MEP HyperCel has also been effective in binding recombinant protein from Escherichia coli. Unlike antibodies, which bind by both affinity and hydrophobic interactions, most recombinant proteins will bind only by hydrophobic interactions and require some lyotropic salt (e.g., ammonium suflate) to effect binding. Because MEP HyperCel is much more hydrophobic than most hydrophobic sorbents used for proteins, lower concentrations of salt can be used.

Major contaminants from E. coli are lipopolysaccharide (LPS), host-cell proteins, and nucleic acid. Those components tend to bind tightly to MEP because they are all hydrophobic and/or negatively charged. LPS is very negatively charged (highly acidic) and very hydrophobic, so it elutes only under very extreme conditions. Thus, as the protein elutes, the LPS stays bound to the column. Nucleic acid is negatively charged, so as the sorbent becomes more positively charged and the protein elutes, the nucleic acid will bind more tightly. Many host-cell proteins also bind tightly to MEP HyperCel. If used as the polishing step following Protein A capture, mixed-mode chromatography with MEP could be used to remove leached protein A, which is acidic.

Two recently introduced sorbents, "HEA" and "PPA" (Pall Life Sciences) are hydrophobic resins that have different ligands (a hexyl group and a phenyl group, respectively) and a positive charge until pH 8 or higher. Adsorption to HEA and PPA is performed at a pH above their isoelectric point (pI) and eluted by decreasing the pH below their pI. As the protein becomes positively charged, they are repelled from the positive charges on the HEA and PPA resins. Proteins can bind to HEA and PPA even below their pI by adding lyotropic salts such as ammonium sulfate. Elution of the protein from the sorbent in this case is effected by lowering the salt concentration. "The selectivity of HEA and PPA differs and is not always predictable, but this adds another tool to the chromatographer's kit," says Lees. "The properties of mixed-mode sorbents not only provide unique selectivities but, by combining features of multiple modes, can reduce the total number of column steps needed."

Libraries of mixed-mode ligands

GE Healthcare Bio-Sciences (Uppsala, Sweden, www.gelifesciences.com) is designing chromatographic ligand libraries following combinatorial chemistry strategies, which take into account a wide range of parameters and interactions to design ligands for mixed-mode applications.

The use of specialized sorbents may reduce the number of process steps. (IMAGE COURTESY OF PALL)

"This approach provides the possibility to tune the performance of your media," says Jean-Luc Maloisel, PhD, staff scientist. "We use it to discover new selectivities and to improve discriminate chromatographic media. GE Healthcare has the traditional media, and we try to introduce new interactions to modulate the performance of these media." The company designs a library of mixed-mode ligands comprising about 50 prototypes to start with. These first, diverse libraries are used in the initial screening step to find a set of interactions that will perform according to a desired specification (e.g., for protein recovery, protein binding, elution).

After the identification of promising candidates, a second-generation library is designed by introducing additional interactions, thereby fine-tuning the modification to the media to reach the final target specification.

"We can screen our libraries of media very rapidly and in different ways using microtiterplate formats. The libraries are not set in stone, and we are always continuously working to diversify them. The libraries we have now may be slightly different tomorrow, so we can design them in different ways," says Maloisel.

Traditional methods are robust, but may not be the optimal solutions in terms of process economy and time to market. "In some cases you really would like to have a wider range of selectivities and have more customized solutions. And that is where the library of multimodal ligands can be useful. However often, people will not take the time and trouble to screen a lot of media. This is where the microtiterplate format allow for a high-throughput process development and that is where you have a chance to win" says Maloisel.

GE has launched multimodal cation exchangers as part of its "Capto" platform. One challenge in designing these media was to develop a medium allowing for binding proteins under high-salt conditions without the need to dilute the feedstock. Using computerized programs, the company made a diverse range of cation exchangers and then screened this range with the condition that the prestandard be salt tolerant. The company also developed anion exchange media for monoclonal antibody capture. "It doesn't replace the other chromatographic media, but it can replace traditional chromatography media lacking the prestandard performance."

UpFront Chromatography (Copenhagen, Denmark, www.upfront-dk.com) also has developed a library of 200 ligands. The general criteria are that these ligands must be nontoxic, simple to manufacture to high purities, stable to harsh cleaning, and stable when immobilized. The performance of the ligands can be modified using various approaches.

Process chromatographers continue to work toward systems that are more economical, offer better selectivity, and lead to greater yields. (IMAGE COURTESY OF PALL)

"It depends on the constituent groups," explains Rob Noel, business development manager at UpFront. "If you take a group of ligands that have a common ionic component such as a carboxyl group, which is negatively charged, then you can change the pI of that carboxyl group depending on its substituents in the ligand."

Mixed-mode ligands can use a combination of aromatic, hydrophobic, ionic, and hydrogen bonding groups. Each combination of chemical groups may have a particular target on the protein surface. Most mixed-mode ligands bind to the surface of protein.

"We try to mimic in some way how a protein–protein interaction would work even though the chemistry is different," says Noel.

Designing a ligand for a protein depends on the characteristics of the protein, and some general rules can be followed. "For example, if the pI is greater than 7, then because most contaminating proteins are usually acidic, then chances are you can quickly find an absorbent that would work," says Noel. "If it's right in the middle in terms of pI, then you would look to see how concentrated it is in your environment. If it stands out in concentration, such as 10 to 100 times more concentrated than any other protein, which in a recombinant system you would hope it is, then that helps."

There is increasing interest in using Protein A and then using a mixed-mode ligand in the next step. Traditionally, the protein A step is followed by two ion exchangers. "Using a mixed-mode ligand is giving an added advantage and effectively trying to replace those two ion exchangers with one step separation. So your whole process is just two steps," says Noel.

Recent improvement in mixed mode chromatography involves finding faster ways of identifying the best and most robust conditions. "People are taking one absorbent and looking at a wide range of conditions to get a fuller picture of how the absorbent is working in their system," says Noel. The practice has become popular in the past four years. "It gives the process engineer an understanding of how the absorbent is working in that process and where the no-go areas are, such as for pH, conductivity, or temperature."

Affinity ligands

Some process chromatographers point out that mixed-mode ligands have a broad range of applications, and thus tend to be unselective and provide relatively modest increases in purity. For this reason, bioscience firms are developing small chemical ligands for capturing specific target proteins (see Figure 2).

Figure 2

For example, ProMetic Biosciences, Ltd. (Cambridge, UK, www.prometic.com) has designed a library comprising tens of thousands of affinity ligands (under its "Mimetic" trade name), which are more specific versions of mixed-mode ligands. Like mixed-mode ligands, these are small, synthetic compounds, but they differ greatly from what are generally understood as mixed-mode ligands because of their precise chemical composition and wide variety of chemical groups used. These affinity compounds are organized such that the orientation and position of chemical groups in a three-dimensional space is fixed. This fixed position enables interaction with specific and complementary groups on the protein, which in turn enables specific capture. The synthetic ligands are designed and developed specifically according to the target protein (e.g, monoclonal antibodies, albumin, glycoproteins, proteases, and plasma proteins).

"In many cases, we have already identified compounds that are effective in purifying particular proteins," says Steve Burton, CEO, ProMetic Biosciences, Ltd. "For example, we have some very good ligands that we have developed and scaled up for albumin purification. Sometimes we have clients approach us with a new protein that maybe has just been discovered. In this case we may not have anything currently developed for that new target, but within our ligand libraries we can usually find something that will bind specifically and can be used to purify it."

The affinity ligands are more complex than those used in ion exchange and HIC. Charge interactions, hydrophobic interactions, hydrogen bonding and van der Waals interactions plus specific three-dimensional geometries are incorporated into the media. "These are not generalized interactions," says Burton. With ion exchange, for example, the interaction is not normally on a specific site on the protein; its interaction is with the protein as a whole. "These small, synthetic ligands, however, are actually interacting with a very specific site on the protein. We engineer the ligands so they have complementary charge and hydrophobic groups, depending on the groups that are present in that specific region of the protein."

As Burton observes, one of the benefits of having small synthetic ligands is they are less expensive to produce than protein-based ligands such as Protein A. In addition, being small entities, they are stable, robust, and can be cleaned effectively with molar sodium hydroxide, for example, so they can be used for hundreds of cycles.

"Just a few years ago there was a lot of skepticism about whether one could actually use these small compounds in bioprocesses," says Burton. "Now people are realizing that there are some real benefits. They really do work well in many instances, and more time is being devoted to investigating their potential."