Specificity and polyspecificity
A binder that sticks to everything will look excellent in a binding assay against its target and then clear from circulation in hours, light up every tissue in an imaging study, and produce cell results that have nothing to do with its epitope. Polyspecificity is one of the standard reasons a molecule dies late, and it is cheap to screen for early.
What drives it
Two properties dominate. The first is a large positively charged patch on the variable domains; positive charge drives binding to cell surfaces, to heparan sulfate, and to nucleic acids, and it correlates with faster clearance. The second is exposed hydrophobic surface on the CDRs, which drives non-specific association and aggregation at the same time.
Both are consequences of the CDRs being selected for binding without any pressure against binding other things, which is exactly what happens in a campaign that only selects positively.
Screening for it
Several assays are in routine use, and they measure related but not identical things.
- Polyspecificity reagent binding. A soluble membrane preparation or a mixture of biotinylated cellular material, incubated with the antibody, read by ELISA or flow. Widely used as a proxy for clearance risk.
- Baculovirus particle ELISA. An old and well-validated proxy that correlates with clearance in several published panels.
- DNA or heparin binding. Cheap, and a direct readout of the charge-driven component.
- Cross-interaction chromatography and self-interaction measurements. These probe the colloidal behavior that shows up later as viscosity and aggregation.
- Binding to an unrelated cell line panel. The most literal version of the question.
None is definitive alone. Running two that probe different mechanisms, typically one charge-driven and one membrane-based, catches most problems.
Fixing it
Polyspecificity often tracks to a small number of residues. Removing one or two arginines or lysines from a positive patch, or replacing an exposed aromatic, can cut non-specific binding substantially with little effect on the target interaction, and the change is testable in a week.
Where the sticky residues are also the binding residues, you have a choice to make with data rather than a rule to apply: measure both the specific and the non-specific binding for a small variant series and pick the trade you can live with.
Building it into selection
Counter-selection during display, negative sorting against an unrelated cell line, and including a polyspecificity screen in the first triage all cost little and prevent a bad molecule from consuming months. A panel that has never been asked what else it binds is a panel with an unknown.
The pattern worth avoiding: ranking a panel on affinity, choosing a lead, spending a quarter on developability and pharmacokinetics, and discovering that the tightest binder was tight partly because it was sticky.