Why can a host cell protein ELISA pass while a problematic host cell protein remains?

Why

Host cell protein ELISAs use polyclonal antibodies raised by immunizing animals with proteins from a null cell line. A 2022 review notes the ELISA gives no information about the identity or amount of individual proteins, that low molecular weight proteins are often less immunogenic and therefore poorly detected, and that insufficient coverage may require orthogonal methods such as mass spectrometry. It also notes that even minute amounts of proteolytic host cell proteins can cleave the product over time, reducing potency and stability.

So a total ELISA value well within the limit does not prove that no harmful protein is present. Troubleshooting starts with a symptom, such as slow fragment growth on stability or particle formation, that suggests an active impurity. LC-MS of the drug substance can identify individual proteins and estimate their levels.

Coverage studies, such as immunoaffinity capture followed by mass spectrometry, show which proteins the ELISA antibodies can see. If a protease or lipase is found, tracking it through each purification step shows where it co-purifies, often by binding the product itself. Fixes include adjusting wash conditions or adding a polishing step.

Guidance sets no universal numerical limit, because risk depends on dose, route, frequency and the particular impurity, as the review states.

How to work it through

1. Link the symptom, such as fragmentation or particles, to a plausible enzymatic or immunogenic cause.
2. Run LC-MS on drug substance to identify individual host cell proteins.
3. Check ELISA antibody coverage of those proteins with a coverage method such as immunoaffinity capture and mass spectrometry.
4. Track the protein through each purification step to see where it co-purifies.
5. Test wash or polishing changes at small scale for clearance.
6. Update the control strategy, possibly with a specific assay for the protein.

Where the answer stops holding

LC-MS sensitivity is limited for very low abundance proteins in a sea of product. Linking a specific protein to a quality effect needs spiking or inhibition experiments. Coverage percentages vary by method, so comparisons across studies are rough.