ADC linker chemistry
The linker decides where the payload is released and what form it is released in. That determines whether neighboring antigen-negative cells are killed, how much payload leaks in circulation, and which normal tissues see a dose. It is the component most often chosen by default and most deserving of thought.
Non-cleavable linkers
A non-cleavable linker relies on the whole conjugate being internalized and trafficked to the lysosome, where the antibody is proteolyzed down to amino acids. The released species is the payload still attached to its linker and to the terminal residue, typically a charged lysine or cysteine adduct.
That charge is the defining property. A charged catabolite does not cross membranes, so it stays in the cell that internalized it. There is no bystander killing, and there is very little systemic exposure to free payload.
This suits a homogeneously expressed target with reliable internalization, and it gives the cleanest safety profile. Ado-trastuzumab emtansine is the reference case. The cost is that any antigen-negative cell in the tumor survives, and solid tumors are rarely homogeneous.
Cleavable linkers, and what each one assumes
Protease-cleavable. Valine-citrulline with a self-immolative PAB spacer is the workhorse, cut by cathepsin B in the lysosome and releasing free, membrane-permeable payload. It assumes lysosomal delivery and adequate cathepsin activity. It was long described as plasma-stable, and it is in human plasma, but mouse plasma carries Ces1C carboxylesterase which cleaves it, so mouse tolerability studies systematically overstate the toxicity you will see in patients unless you use a Ces1C knockout.
Glucuronide. Beta-glucuronidase cleaves a glucuronide linker in the lysosome. The linker is markedly more hydrophilic than val-cit, which helps with the aggregation that high-DAR hydrophobic payloads cause, and it avoids the mouse esterase problem.
Disulfide. Release depends on the intracellular glutathione gradient, with steric hindrance around the disulfide tuning the rate. The assumption is a large enough gradient between plasma and cytosol, which holds, though the margin is narrower than the mechanism suggests and hindered disulfides are needed to get there.
Acid-labile hydrazone. Cleaves at endosomal and lysosomal pH. This was the gemtuzumab chemistry and it taught the field its lesson: hydrazones hydrolyze slowly at pH 7.4 as well, so payload leaks in circulation over days. Largely superseded.
Bystander killing is a decision
A cleavable linker releasing a neutral, permeable payload means that payload diffuses out of the cell that took it up and kills neighbors. In a heterogeneous solid tumor that is often the difference between activity and none, because the antigen-negative fraction is otherwise untouched.
It is also the mechanism by which normal cells adjacent to low-level antigen expression are killed. Interstitial lung disease with several ADCs and ocular toxicity with others track to exactly this.
So the choice follows the biology. Heterogeneous antigen expression in a solid tumor argues for bystander activity. A uniformly expressing hematological target with a narrow therapeutic window argues against.
Hydrophobicity is the practical constraint
Most payloads are hydrophobic. Attaching four or eight of them to an antibody creates surface hydrophobic patches, and those drive aggregation, faster clearance and higher liver uptake. The relationship between DAR and clearance is steep above DAR 4 for hydrophobic payload-linker combinations.
The linker is the lever. Inserting PEG units, using a hydrophilic cleavable motif such as the glucuronide, or placing a charged sulfonate in the spacer all mask the payload and restore pharmacokinetics closer to the naked antibody. This is why modern high-DAR conjugates with eight camptothecin payloads are viable where an eight-load auristatin on a plain val-cit would not be.
Measure it rather than assume it. Hydrophobic interaction chromatography retention time correlates well with in vivo clearance, and it is a one-day experiment that predicts a six-week study.