Almost every monoclonal antibody in development today follows the same route: capture on an affinity ligand, a low-pH hold for virus inactivation, one or two polishing steps to remove what capture left behind, virus filtration, and formulation — with conditioning and depth filtration between them. The route is so standardised that it is called the platform process, and its main virtue is that it works for most molecules without redevelopment. This page walks through it, names the decision at each step, and points to where each one is covered in detail.
The route at a glance
| Step | What it does | The decision to make |
|---|---|---|
| Capture — affinity chromatography | Binds the antibody from clarified harvest, removes most host-cell protein, DNA and medium components in one operation | Which ligand: Protein A for most human IgG, Protein G where Protein A binds weakly |
| Elution and low-pH hold | Releases the product at pH 3.0–3.6. The eluate is then titrated to a validated inactivation pH, typically 3.6 or below, and held 30–60 minutes under temperature control | Buffer species and how fast the pool is neutralised — the largest driver of aggregation |
| Polishing | Removes aggregate, residual host-cell protein, DNA and leached ligand | Which mode: cation exchange or multimodal for aggregate, anion exchange in flow-through as the second polish |
| Virus filtration | Removes small non-enveloped viruses — parvoviruses are 18–26 nm, the antibody 8–12 nm — which the low-pH hold does not touch | Filter grade and where it sits; it is the second, orthogonal clearance step and not optional |
| Formulation | Buffer exchange and concentration into the final formulation | Usually ultrafiltration/diafiltration; outside the scope of this page |
How does affinity capture work?
Protein A binds the Fc region of immunoglobulin G through a largely hydrophobic contact at the CH2–CH3 interface. Because that contact is specific to the Fc and almost nothing else in a serum-free harvest carries it, a single capture step removes the overwhelming majority of host-cell protein, DNA and medium components. Almost, not nothing: some host-cell proteins bind non-specifically or travel with the product, which is why wash design is a development topic of its own — and serum in the medium brings its own IgG, which the ligand cannot distinguish from the product at all — a purification factor no other single operation in the train comes close to. That is the reason the platform is built around it despite the cost of the medium.
The full cycle — equilibrate, load, wash, elute, strip, clean, re-equilibrate — is set out step by step, with buffer table, in how to run a Protein A capture cycle. MonoCore™ Protein A runs that cycle as a monolith column supplied as a drop-in capsule for standard FPLC systems, in 1.8 mL and 5.4 mL with UNF 10/32 connections.
Where Protein A is the wrong ligand. Binding strength is a property of the subclass, not of the format. Mouse IgG1 — the subclass most hybridoma lines produce — binds Protein A weakly, which is why Protein G is the standard capture ligand for hybridoma and much veterinary work. The subclass table and a capture protocol for that case are in antibody purification from hybridoma supernatant.
Why is the elution step the difficult one?
Because the chemistry that releases the product is also the chemistry the product has to survive. Protein A releases its target when the histidines in the binding interface are protonated, which happens around pH 3.0 to 3.6 — conditions far from where an antibody is stable. For a robust IgG1 that is a few minutes of discomfort. For a bispecific with an asymmetric interface, an acid-labile conjugate or a fragile fusion protein, it is the step that decides whether the molecule survives the process.
Three pages cover this in detail: which elution buffer to use for the choice between glycine, citrate and acetate; why low-pH elution fails on sensitive modalities for what actually breaks; and how to elute from Protein A without acid for the nine routes away from pH 3 and which three genuinely remove the acid.
One point is worth making here because it is so often missed: in a process that runs a low-pH virus-inactivation hold, the hold is usually the longer acid exposure, not the elution. Measuring aggregate directly in the eluate and again after the hold costs one analytical run and tells you which of the two to work on.
How is virus clearance handled?
By two orthogonal virus clearance steps, not one — this is where a short description of the platform most often goes wrong. The low-pH hold after capture inactivates enveloped viruses and does nothing against small non-enveloped ones. Those are removed by virus filtration, a 20 nm-class filter that retains parvoviruses at 18–26 nm while the antibody at 8–12 nm passes through. Anion exchange in flow-through adds a third, independent mechanism, binding virus along with DNA and acidic host-cell protein.
Three mechanisms that fail in different ways is the whole point: inactivation by chemistry, removal by size, removal by charge. A process relying on the acid hold alone would have no clearance for parvovirus at all.
Two further operations belong in any realistic account of the route, even though neither is chromatography. The neutralised virus-inactivation pool regularly carries precipitate, which is removed by depth filtration before the next column. And between steps there is conditioning — dilution, pH and conductivity adjustment — which in practice often costs more volume and time than the chromatography it prepares for.
What does polishing have to remove?
Four things. Aggregate is usually named as the hardest, but a host-cell protein that travels with the product can be just as stubborn — the order depends on the molecule more than on the class of impurity.
- Aggregate. Dimers and higher species, often formed during the acid step itself. Usually removed by cation exchange, where aggregate and monomer differ in net charge, or on a multimodal medium; hydrophobic interaction separates on exposed hydrophobic surface instead and is the more reliable route where the charge difference is small — the comparison is in aggregate removal on a HIC monolith.
- Residual host-cell protein. Capture removes most of it; polishing removes the fraction that co-elutes, often the part that associates with the product itself.
- Host-cell DNA. Usually handled in flow-through mode on an anion exchanger, where DNA, acidic host-cell proteins, leached ligand and virus bind while the antibody passes — which is why this step is close to universal.
- Leached ligand. Protein A that has come off the medium travels bound to the Fc and has to be cleared downstream. Rising leaching over a lifetime is one of the end-of-life signals, alongside falling capacity and yield — see regeneration, CIP and service life.
Where the route changes for other formats
The platform is built around full-length human IgG1. Three deviations come up often enough to plan for.
Hybridoma and veterinary antibodies. Different subclass, therefore often a different ligand, and serum in the medium brings its own IgG that the ligand cannot distinguish from the product. Covered in the hybridoma page.
Bispecific antibodies. The asymmetric interface is usually the least stable part of the molecule and relaxes before the rest does, so the acid step is a sharper constraint. Mispaired species are an assembly problem, not a purification one, but polishing has to separate them.
Fc-fusion proteins and ADCs. The fused partner or the linker chemistry, not the Fc, sets the limit. An acid-labile linker does not care how briefly it sees the acid — for those, a genuinely non-acidic elution is the only route.
Which format: packed column, membrane or monolith?
The chemistry decides what binds; the format decides how fast and how robustly. A packed bed of porous beads is the established format with by far the longest track record in licensed processes; transport there is by diffusion into the pores, which is why it is run at residence times of minutes. Membrane adsorbers and monolith columns are convective: the liquid carries the target past the ligands, so the same method runs at several bed volumes per minute rather than a fraction of one.
What that is worth depends on whether time is your constraint. It matters for multi-cycle capture of a large batch and for process development screening where each experiment costs a day. It also matters for acid-sensitive material that does not undergo a low-pH hold — research-grade antibodies, fragile fusions, acid-labile conjugates — where the elution is the only acid exposure there is. Where a 30–60 minute hold follows, shortening an elution of a few minutes changes little, as the section above says. It matters very little if the step runs once a week and nothing waits for it. The trade is set out in monolith vs. membrane chromatography, and the arithmetic for your own method in the run planner.
Capacity is a separate question from speed and belongs to the datasheet of the specific product, together with the conditions it was measured under — what those conditions mean is in how to read a dynamic binding capacity figure.
When something goes wrong
The commonest failures in a capture step are low yield, product in the flow-through, a strip peak that is larger than it should be, and rising back pressure. The first move is always the same: find out where the product went before changing anything. The systematic version is in Protein A column troubleshooting.
Frequently asked questions
How do you purify a monoclonal antibody?
In five unit operations, plus the conditioning between them. Capture on an affinity ligand — Protein A for most human IgG — which removes the majority of host-cell protein, DNA and medium components in one operation. Elution at pH 3.0–3.6, usually followed by a low-pH hold that inactivates enveloped virus. One or two polishing steps to remove aggregate, residual host-cell protein, DNA and leached ligand, typically by cation exchange and anion exchange in flow-through. Then buffer exchange and concentration into the final formulation.
Why is Protein A used for antibody purification?
Because it binds the Fc region specifically, and nothing else in a cell culture harvest carries that region. A single Protein A step therefore achieves a purification factor that no other single operation in the train matches — purity typically better than 95 % by SDS-PAGE and a host-cell-protein reduction of roughly three orders of magnitude, on the figures commonly reported for the platform. That still leaves an eluate in the region of 10² to 10⁴ ppm host-cell protein against a target usually below 100 ppm — which is why polishing is not optional, however good the capture step looks. That is what justifies the cost of the medium and why the platform process is built around it.
What is the difference between capture and polishing?
Capture takes the product out of a complex, dilute feed and concentrates it, accepting whatever co-purifies. Polishing starts from an already pure solution and removes the remaining specific impurities — aggregate, residual host-cell protein, DNA, leached ligand. Capture is about selectivity and capacity; polishing is about resolution.
Does antibody purification always need Protein A?
No, but the alternatives cost something. Protein G is used where Protein A binds weakly, notably mouse IgG1. Mixed-mode and ion exchange capture avoid the affinity medium entirely, at the price of much lower selectivity — you give up the largest single purification factor in the train and the polishing steps have to absorb the difference. For an antibody fragment without an Fc region, Protein A is not an option at all.
What causes aggregation during antibody purification?
Mostly the acid. Time at low pH during elution and during the virus-inactivation hold is the dominant cause, and the damage rises with both how low the pH goes and how long it lasts. The neutralisation path matters as much as the elution itself: a pool that crosses its own isoelectric range slowly will precipitate. Pre-loading the neutralisation buffer so each fraction is neutralised as it lands is the cheapest fix there is.
How much antibody can one column purify?
The binding capacity of the medium times the bed volume, loaded to a fraction of that figure rather than to the last milligram — how large a margin depends on how well characterised the step is and how variable the feed. That capacity figure applies at a specific residence time, feed and bed height, so it does not transfer unchanged between processes. If the batch exceeds what one cycle holds, the batch is run over several cycles — which is a design choice in itself, covered in rapid cycling chromatography.
Still an open question?
Tell us the molecule, the feed volume and the system you run it on, and we will say whether a MonoCore™ capsule is the right starting point — or whether it is not.