Bispecific antibodies

A conventional antibody works by occupancy: it sits on a target and stops something from happening. A bispecific antibody has two different binding sites on one molecule, and that changes the category of the tool. Instead of blocking, it can hold two entities together that would otherwise never meet — and proximity, in biology, is frequently the whole signal.

Forced proximity as a drug mechanism

The clearest case is the T-cell engager. One arm binds CD3ε on a T cell, the other a tumour surface antigen such as CD19, BCMA, GPRC5D or DLL3. Crosslinking pulls the two membranes into an immunological synapse and triggers CD3 signalling directly, bypassing the normal requirement for peptide-MHC recognition and for costimulation. The T cell then kills by the ordinary perforin-granzyme route. Nothing about the patient’s own T-cell repertoire needs to recognise the tumour, which is why the approach works in tumours with no useful neoantigens and why it can be supplied off the shelf rather than manufactured per patient.

The second case is purely biochemical. Emicizumab bridges activated factor IX and factor X, holding them in the geometry that factor VIII would normally impose. It has no sequence or structural relation to factor VIII, so it works in haemophilia A patients who have developed inhibitors against factor VIII replacement. Here the drug is not immunological at all; it is a molecular clamp restoring a reaction rate.

Where the difficulty is

Chain pairing is the manufacturing constraint. An IgG assembled from two different heavy chains and two different light chains can combine into a large number of species, most of them inactive or monospecific. The engineering answers — complementary knob-into-hole mutations in the CH3 interface, charge-pair steering, common light chains, controlled Fab-arm exchange, or dropping the Fc altogether for tandem single-chain formats — all exist to suppress mispairing. Format is not cosmetic: blinatumomab, a roughly 55 kDa tandem scFv, has no Fc and therefore no FcRn recycling, giving a half-life of about two hours and requiring continuous intravenous infusion, while Fc-bearing engagers dose intermittently.

Affinity has a counterintuitive optimum. Engager potency follows a bell-shaped curve: at high concentrations, both arms saturate their targets separately and the crosslink that does the work becomes less probable — the classic hook effect. Excessively high CD3 affinity also biases distribution toward T-cell-rich lymphoid tissue and away from the tumour, so many programmes deliberately de-tune the CD3 arm.

The ceiling

Because activation is forced rather than negotiated, it is not self-limiting. Bulk T-cell activation releases IL-6, IFN-γ and TNF, producing cytokine release syndrome, and CD3-driven engagers also cause neurotoxicity whose mechanism is still not fully resolved. The field’s response is exposure shaping rather than a change of mechanism: step-up priming doses, mandatory monitoring periods, and tocilizumab availability. Target selection carries the same weight, since T-cell killing has no expression threshold — a small amount of antigen on healthy tissue is enough to be attacked, which is the reason on-target off-tumour toxicity, not potency, limits most solid-tumour engager programmes.

See also the technology article at /technology/bispecific-antibodies/.