ADC payload classes
The payload has to be potent enough to work at the amount an antibody can deliver. A tumor cell might internalize a few hundred thousand antibody molecules, which puts the intracellular payload concentration in the nanomolar range, so anything with a conventional chemotherapeutic potency does nothing. That single constraint explains why the classes in use are the ones they are.
Tubulin inhibitors
Auristatins are synthetic dolastatin analogs. MMAE is the permeable version and gives bystander killing when released from a cleavable linker. MMAF carries a charged C-terminal phenylalanine, does not cross membranes and is used with non-cleavable linkers for contained killing. Both are in the sub-nanomolar range against dividing cells.
Maytansinoids bind at the vinca site. DM1 and DM4 differ mainly in the linker chemistry and in whether the released catabolite is permeable.
The class limitation is mechanistic. Tubulin inhibitors only kill cells that divide, and the quiescent fraction in a solid tumor is substantial. Peripheral neuropathy is the characteristic dose-limiting toxicity, and it is cumulative and slow to reverse.
Topoisomerase I inhibitors
Camptothecin derivatives, deruxtecan and SN-38 being the ones in approved products, have shifted the field. They are less potent per molecule than auristatins, which sounds like a disadvantage and is not.
Lower potency permits a higher DAR. Deruxtecan conjugates run at DAR 8 where an auristatin conjugate sits at 2 to 4, so the delivered payload per binding event is higher and the therapeutic window is set by a different toxicity. The payload is also membrane-permeable and gives strong bystander killing, which is why these conjugates work against tumors with low and heterogeneous antigen expression where earlier generations failed.
They kill non-dividing cells as well, since topoisomerase I is active in transcription rather than only replication.
The characteristic toxicities are myelosuppression and interstitial lung disease. The pneumonitis is class-associated, it can be fatal, and it is the reason these products carry monitoring requirements.
DNA damaging agents
Pyrrolobenzodiazepine dimers crosslink DNA in the minor groove at picomolar potency. Calicheamicins cause double-strand breaks. Duocarmycins alkylate adenine.
Extreme potency means very low DAR, often 1 to 2, and a correspondingly narrow window. Several PBD programs have been discontinued for delayed and cumulative toxicity, including hepatotoxicity and effusions, that did not appear until late in trials. The class is not dead, and it is now reserved for targets where the expression differential is unusually clean.
Emerging classes
RNA polymerase II inhibitors such as amanitin kill non-dividing cells at very low concentration. Immune agonists, TLR7/8 and STING ligands, change the objective entirely: the payload recruits an immune response rather than killing the cell directly, which changes both the potency requirement and the relevance of internalization.
Choosing, and the tradeoff to hold in mind
Potency and DAR trade against each other, and the product of the two is what reaches the cell. A more potent payload does not mean a better conjugate; it means a lower DAR, a narrower window and greater sensitivity to any off-target uptake.
Match the mechanism to the tumor. A slowly proliferating tumor is a poor match for a tubulin inhibitor no matter how potent.
Match the release chemistry to the antigen distribution. A permeable payload plus a cleavable linker gives bystander killing, which heterogeneous expression needs and uniform expression does not.
Expect the payload to set the toxicity. Most ADC dose-limiting toxicities are payload-driven and off-target, arising from uptake by normal tissue independent of the antigen, which is why conjugates sharing a payload share a toxicity profile across completely different targets.