ADC aggregation and hydrophobicity
Conjugation makes an antibody less soluble and less stable. Payloads are hydrophobic small molecules, and attaching several to a protein surface creates patches that were not there before. Aggregation rates go up, clearance goes up, and the effect scales steeply with DAR.
Two mechanisms, not one
Surface hydrophobicity is the obvious one. Exposed payload drives self-association, and the relationship with DAR is nonlinear: DAR 2 is often close to the naked antibody, DAR 8 with a hydrophobic auristatin can be an order of magnitude worse.
Structural destabilization is the one people miss. Cysteine conjugation requires reducing interchain disulfides, and those bonds are not replaced. The conjugate is held together by the non-covalent interface between heavy and light chains, which is strong but no longer reinforced. Under thermal or mechanical stress the chains dissociate, and half-antibody and free light chain appear on SDS-PAGE under non-reducing conditions.
Lysine conjugation does not do this, because no disulfide is broken. Bridging chemistries that re-staple the reduced pair recover it deliberately.
Measuring hydrophobicity, and why it predicts the animal study
HIC retention time is the workhorse. It is the same method used for DAR, read differently: the retention of the whole distribution, and particularly of the high-DAR tail, tracks with in vivo clearance closely enough to use as a screening filter.
That correlation is the single most useful thing in ADC developability. A six-week pharmacokinetic study in mice can be anticipated by a one-day chromatography run, and a linker-payload combination that retains far longer than its comparator will clear faster, accumulate in liver and show a worse therapeutic index.
Standardize the gradient and always run the unconjugated antibody in the same sequence as the anchor. Absolute retention is method-dependent, and the shift against the naked antibody is the transferable number.
The rest of the panel
Size exclusion chromatography for soluble aggregate, run against the naked antibody. Report monomer percent and total recovered mass, because insoluble aggregate does not elute and a purity-only readout will score a sample that lost material to the guard column as clean.
Non-reducing SDS-PAGE or capillary electrophoresis for chain dissociation. This is where half-antibody shows up, and it is specific to cysteine conjugates.
Differential scanning calorimetry or a thermal shift assay for domain stability. The CH2 domain is usually the first to unfold, and conjugation there lowers its transition temperature measurably. A drop of several degrees against the parent predicts trouble on storage.
Subvisible particle counting once a formulation candidate exists. Particles in the two to ten micron range are the immunogenicity concern and are invisible to SEC.
Recovering it
Linker hydrophilicity is the most effective lever. PEG spacers, a glucuronide in place of val-cit, or a charged sulfonate in the linker all mask the payload without touching the antibody, and they are the reason DAR 8 conjugates are viable now.
Site selection matters for engineered-cysteine formats. Some positions bury the payload against the protein surface and some display it, and the difference in HIC retention between two sites carrying identical payload can be substantial.
Formulation helps at the margin. A nonionic surfactant addresses interfacial aggregation, sucrose or trehalose stabilizes against thermal and freeze stress, and histidine around pH 6 is the usual starting buffer. Formulation will not rescue a conjugate that is intrinsically too hydrophobic, and trying is a common way to spend six months.
Purification removes the tail. Preparative HIC can strip the DAR 8 species from a DAR 4 average, which improves pharmacokinetics directly and costs yield. It is a legitimate manufacturing choice rather than an admission of failure, though it is better to not make the species in the first place.