"Do I have to filter it first?" is the question we are asked more often than any other, and the honest answer has two parts. Yes — every chromatography step wants a clarified feed. But how far you have to go is not a fixed rule, and the usual failure is not too little filtration. It is carrying over the filter train from the last project without asking whether this feed needs it.
This page is about that decision: what each stage removes, what it costs, and how to tell when you have gone far enough.
What is actually in the feed
The word "feed" covers materials that have almost nothing in common. What has to come out depends entirely on where it came from.
| Feed | What is in it | What clarification has to achieve |
|---|---|---|
| Cell culture supernatant, low density | Cells, some debris, medium components | Remove cells and the coarse fraction; often centrifugation plus one filter stage |
| High-density or late-harvest supernatant | More debris, lysed cells, DNA released from them | A depth filtration train; DNA raises viscosity and blinds filters faster than expected |
| Hybridoma supernatant with serum | Lipid, serum protein, aggregates | Lipid is the problem, not the particle count — it passes filters and fouls the inlet |
| Lysate after cell disruption | Membrane fragments, host DNA, the whole intracellular content | A multi-stage train; the debris spans a wide size range and no single filter catches it |
| Nuclease-treated lysate | The above, minus most of the DNA | Easier than untreated lysate, but the nuclease is now an impurity of its own |
The three tools, and what each one does
Centrifugation removes what is heavy: cells and the coarse debris fraction. It is cheap at scale and takes the bulk out, but it leaves the fines behind, and a continuous centrifuge at high shear can create new fines by breaking cells that were intact.
Depth filtration is the workhorse, and it is often misunderstood. A depth filter is not a sieve with one pore size — it is a graded fibre bed that retains particles through its whole thickness, by size and by adsorption. That is why depth filters are specified by area and by capacity rather than by a single micron figure, and why two filters with the same nominal rating behave differently on the same feed.
Membrane filtration at 0.2 or 0.45 µm is absolute retention: what is larger does not pass. It is a guard step, not a clarification step. Running it on a feed that has not been properly clarified first is the classic way to blind an expensive filter in minutes.
How far down the train do you have to go?
Far enough to protect the column, and no further. That sounds evasive, so here is the reasoning behind it.
Every stage costs product. Depth filters adsorb — that is part of how they work — and the loss is not negligible, least of all for large particles such as viral vectors, where it is routinely reported as a significant contributor to overall yield loss. A stage you add "to be safe" is a stage that takes yield with it every single run.
The second consideration is the medium. The same feed can need a 0.2 µm guard filter on one format and run without it on another, which is why the filter train is a property of the pair — feed and medium — rather than of the feed alone. Ask the supplier what their format expects rather than carrying a train across.
The practical approach: start from what the medium supplier states for their format, run it, and watch the pressure. A stable pressure across a cycle means the train is sufficient. A pressure that climbs run after run means it is not — or that the feed is being held too long before loading.
Where the yield goes
| Stage | Loss mechanism | What reduces it |
|---|---|---|
| Centrifugation | Product in the pellet, shear on fragile particles | Gentler conditions; check the pellet rather than assume |
| Depth filtration | Adsorption to the fibre bed | Fewer stages, correctly sized area, and a flush where the product tolerates it |
| 0.2 µm guard filter | Adsorption, and retention of aggregate that is product | Only where the medium needs it |
| Hold time before loading | Precipitation and aggregation in the clarified feed | Load promptly; a feed that has stood is a different feed |
Signs that clarification was insufficient
- Pressure rises within a single run. Something is accumulating at the inlet. This is the clearest signal and the easiest to miss if only the end-of-run pressure is recorded.
- Pressure rises across cycles and does not return after cleaning. Material is being fixed on the medium rather than washed off — often because a caustic step fixed what the wash left behind.
- Yield falls over a campaign while purity stays constant. Deposits are blocking access to the medium, and what the feed cannot reach it cannot bind to.
- The first run is fine and the fifth is not. Feed that was held and has begun to precipitate behaves differently from the same feed when fresh.
What these mechanisms are, and how to tell fouling from an overloaded step, is set out in why chromatography media foul.
Where MonoCore™ fits
The feed still has to be clarified — there is no format that removes that step, and we do not claim one. What differs is how much margin sits between the clarified feed and the narrowest part of the flow path. MonoCore™ ships with 15 µm channels, which is wide relative to the gaps in a packed bed, and that margin is what a difficult feed uses up.
In practice that usually means the 0.2 µm guard step matters less than it would on a packed bed. It does not mean an unclarified feed will run. Cells and coarse debris block anything, and a feed that has been standing will cause trouble whatever it flows into.
Frequently asked questions
Do I need to filter the feed before loading?
Yes. Every chromatography step wants a clarified feed — the question is how far to go, not whether. Centrifugation followed by depth filtration is the usual route. Whether you add a 0.2 or 0.45 µm step immediately before the column depends on the medium: it matters most where the narrowest part of the flow path is the gap between packed particles.
What is depth filtration?
Filtration through a graded fibre bed rather than a single membrane surface. Particles are retained throughout the thickness of the filter, by size and by adsorption, which is why a depth filter holds far more solids than a membrane of the same area before it blinds. It is specified by area and capacity rather than by one pore size, and two filters with the same nominal rating can behave differently on the same feed.
Depth filtration or centrifugation — which comes first?
Centrifugation first where the solids load is high: it removes the bulk cheaply and leaves the filter train to handle the fines. For a clean, low-density supernatant, filtration alone is often enough. At small scale centrifugation is frequently the only step that happens, which is fine until the harvest gets denser.
Is a 0.2 µm filter always necessary before the column?
No, and it is not free. It is a guard step, and whether the guard is worth its yield cost depends on what it is guarding. Where the narrowest part of the flow path is the gap between packed particles, it earns its place. On an open channel structure there is more margin, though a feed carrying lipid or precipitate will still cause trouble.
How much product do I lose during clarification?
Enough to be worth measuring rather than assuming. Depth filters adsorb as part of how they work, and for large particles such as viral vectors the effect is pronounced. The number depends on the filter chemistry, the area and the product — which is why every added stage should have a reason beyond caution.
My pressure rises but the feed was filtered. What now?
Check when it rises. Within a run points to something accumulating at the inlet; across cycles that does not clear after cleaning points to material being fixed on the medium. Also check how long the feed stood between clarification and loading — a clarified feed that has begun to precipitate is no longer clarified.
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.