A team builds an antibody-drug conjugate. Their antibody is exquisitely specific — in culture, antigen-positive cells die and antigen-negative cells survive. Confident, they load eight drug molecules onto each antibody. In animals, two things go wrong at once. The conjugate is cleared from blood in hours, and the animals develop liver and bone-marrow toxicity in tissues that do not express the target at all. The linker was too unstable. It leaked a potent cytotoxin into the bloodstream, dosing the whole body with naked chemotherapy. The antibody's specificity was never the problem. The delivery chain had five places to fail, and they had broken two of them.

An antibody-drug conjugate is a three-part molecule, and each part has one job. The antibody is the targeting module — it determines which cell gets bound. That is the only step the antibody controls. The payload is the warhead — far too toxic to give as a conventional drug, killing cells at subnanomolar concentrations, hundreds of times more potent than standard chemotherapy. And the linker is the chemistry tethering them together. The linker is where selectivity actually lives. A linker that releases payload inside the target cell concentrates killing there. A linker that releases payload in the bloodstream turns the antibody into a vehicle for dispersing chemotherapy systemically.

Linkers are classified by how they release payload. Cleavable linkers are cut by conditions elevated inside cells relative to plasma — the endosome and lysosome are more acidic, lysosomal enzymes like cathepsins cleave specific peptide sequences, and glutathione concentrations are a hundred to a thousand times higher inside cells than outside. In principle, a cleavable linker releases payload only after internalization. In practice, they are not perfectly stable in plasma, and premature release determines off-target toxicity. Non-cleavable linkers hold payload through bonds that resist acid, enzymes, and reductants. Payload is freed only when the antibody itself is fully degraded inside the lysosome — released still attached to an amino acid fragment. That fragment is typically charged and hydrophilic. It cannot cross cell membranes.

The bystander effect is what happens when freed payload can cross cell membranes and enter neighboring cells — including cells that never expressed the target antigen and never bound the ADC. The mechanism: an ADC binds an antigen-positive cell, is internalized, the linker is cleaved, and a membrane-permeable payload is released into the cytoplasm. Some of that payload diffuses across the plasma membrane into adjacent cells, killing them even though they expressed no antigen. The consequence of this mechanism was one of the most significant therapeutic advances in breast cancer of the last decade. And the bystander effect is double-edged — the same membrane permeability enables both therapeutic spread in the tumor and toxicity from payload released anywhere, including normal tissues.

The tumor architecture changes which ADC properties are decisive. In a HER2-low tumor, antigen-positive cells are sparse and interspersed with antigen-negative cells. T-DM1 reaches the antigen-positive cells and leaves the antigen-negative majority untouched. T-DXd, with its membrane-permeable payload, kills the HER2-positive cell it enters and the HER2-negative cells around it — spreading killing from a small number of entry points across the entire patch. T-DXd is the correct choice for HER2-low disease, and the decisive variable is the linker-payload combination enabling bystander killing. In a uniformly HER2-high tumor, the calculation shifts. Bystander effect matters less when every cell is antigen-positive and T-DM1 can reach essentially all of them directly. Other factors now drive the comparison: payload class and potency, DAR and clearance, and toxicity profiles — specifically T-DXd's known risk of interstitial lung disease and pneumonitis.

The DAR optimum also depends on where payload attaches to the antibody. Conventional conjugation attaches payload at random surface lysines — producing a heterogeneous mixture of molecules with different loading at different positions. Some subpopulations in the mixture are more aggregation-prone than others, and those are cleared preferentially, pulling the effective DAR of the circulating drug below what was intended. Newer site-specific conjugation methods attach payload at defined positions on the antibody, so every molecule has the same loading at the same location. The aggregation behavior of the uniform population improves, and in principle this allows usable DAR to be pushed higher without the clearance penalty. Whether site-specific conjugation genuinely breaks the DAR trade-off or merely relocates it to other variables is still being worked out.

The most successful ADC in oncology by any measure is T-DXd, and understanding why requires holding five things at once: the antibody that targets HER2, the high DAR that loads enough payload per molecule, the cleavable linker that releases it in the lysosome, the membrane-permeable payload that enables bystander killing, and the pneumonitis risk that membrane permeability also creates. Remove any one of those five and you get a different drug with different clinical behavior. The antibody is the part that gets named in the drug's name. It is not the part that explains its activity in HER2-low disease. That is the whole lesson. Trastuzumab alone can't treat HER2-low breast cancer. T-DM1 can't either. T-DXd can — and the difference is entirely downstream of binding.

Here is the chapter's claim: an ADC's selectivity and clinical behavior are governed downstream of antibody binding — by linker stability, DAR, and payload membrane permeability — so that the antibody alone does not determine the outcome. This is the discipline the chapter asks for: when evaluating an ADC, trace the delivery chain. Ask which step the design improvement actually addresses. Ask which normal tissue shares a dependency with the target tissue at any step where payload escapes. Ask whether the bystander effect is a feature or a liability in the specific tumor architecture being treated. The cleanest finding that would revise this claim: a head-to-head clinical comparison showing antibody affinity dominates outcome while linker and payload design are negligible. To date the evidence runs the other way.