Comparing Cleavable and Noncleavable ADC Linkers

Why Linker Class Changes the Model

Two ADCs can deliver the same nominal payload yet produce different intracellular chemical species. With a cleavable linker, intracellular linker cleavage can release payload directly, often as the unconjugated payload if the linker includes an appropriate self-immolative element. With a noncleavable linker, the antibody is generally degraded in lysosomes and the payload is released as part of a linker- and amino-acid-containing catabolite. The distinction matters because product formation, intracellular retention, membrane passage, and pharmacological activity need not be the same.

A useful comparison therefore holds delivery conditions as similar as possible and changes the processing and product properties that follow from linker design. The earlier material on ADC intracellular release and bystander killing supplies the relevant mechanisms; here they are brought together in a matched comparison. “Matched” does not mean that the two linkers are assumed to have identical kinetics or products. It means that differences in those features are represented explicitly rather than hidden in an unexplained difference in response.

The comparison is conditional on its matching assumptions. If the ADCs have different drug-to-antibody ratio (DAR) distributions, target affinity, uptake, plasma stability, or intracellular trafficking, those differences can also alter exposure. They should either be measured and represented or controlled experimentally. A model that assigns every response difference to linker chemistry without checking these factors does not isolate a linker effect.

A Minimal Matched Processing Model

Consider an intracellular pulse of ADC carrying \(Q_0\) payload equivalents per cell at time zero. Let \(Q(t)\) be the amount of payload equivalents still attached to ADC-derived material and awaiting the linker-dependent processing step. For a simple first-order model, \(k_{\mathrm{rel}}\) is the processing rate constant: it is \(k_{\mathrm{cl}}\) for direct cleavage in the cleavable-linker model and \(k_{\mathrm{cat}}\) for lysosomal catabolism in the noncleavable-linker model. Both have units of inverse time.

The corresponding product-formation rate is \(k_{\mathrm{rel}}Q(t)\). In the cleavable case, denote the released payload by \(P\); in the noncleavable case, denote the payload-containing catabolite by \(M\). The model does not require \(P\) and \(M\) to be chemically or pharmacologically equivalent. If product loss from the intracellular compartment is approximated as first order, with rate constant \(k_{\mathrm{loss}}\), then a generic product amount \(Y\) obeys:

This representation assumes a defined intracellular ADC-derived pool, first-order processing, and a single lumped product compartment. It also assumes that payload equivalents are conserved during the processing step: one equivalent in \(Q\) yields one equivalent of \(P\) or \(M\). If processing is incomplete, produces multiple products, or occurs through distinct endosomal and lysosomal steps, the states and branching fractions must be expanded. The earlier material on linker cleavage and intracellular release gives the basis for deciding when such detail is needed.

Deriving Release Over Time

Separating variables in the equation for \(Q\) gives \(dQ/Q=-k_{\mathrm{rel}}dt\). Integrating from zero to \(t\), then applying \(Q(0)=Q_0\), yields:

The amount processed by time \(t\) is the initial amount minus the amount still awaiting processing. Thus the cumulative processed fraction is \(1-e^{-k_{\mathrm{rel}}t}\). This is a fraction of payload equivalents processed, not necessarily the fraction that remains in the cell as active free payload: product loss and chemical activity are separate parts of the model.

If \(k_{\mathrm{loss}}\ne k_{\mathrm{rel}}\), the solution for the intracellular product amount follows by integrating its formation and loss:

This solution makes the timing distinction explicit. Product can peak and then decline even while cumulative processing continues, because product is also being lost. When \(k_{\mathrm{loss}}=k_{\mathrm{rel}}\), the limiting solution is \(Y(t)=Q_0 k_{\mathrm{rel}}t e^{-k_{\mathrm{rel}}t}\); the apparent division by zero in the preceding expression is removable, not a biological singularity.

Predicting Products and Their Consequences

Product identity follows from the chemistry and processing pathway, not merely from the label “cleavable” or “noncleavable.” A cleavable linker can release unconjugated payload if cleavage and any self-immolation steps produce that species. Depending on linker design and cleavage completeness, the product may instead include a linker-payload intermediate. A noncleavable ADC typically yields a payload-bearing catabolite in which payload remains attached to a linker remnant and an amino acid from antibody degradation. Depending on the conjugation site and chemistry, examples can include cysteine-linker-payload or lysine-linker-payload catabolites. The actual product should be specified from the ADC’s chemical structure and, where possible, confirmed analytically.

These products may differ in charge, polarity, molecular size, membrane permeability, intracellular retention, and ability to bind the pharmacological target. A catabolite is not automatically inactive: if it retains a potent payload pharmacophore and reaches the relevant intracellular target, it may contribute to effect. Conversely, an unconjugated payload is not automatically more effective in the cell if it is rapidly lost or degraded. The model should assign each product its own loss, permeability, and potency parameters when the data support that distinction.

How Product Permeability Shapes Bystander Exposure

A potential bystander effect requires payload-containing material to leave a target-positive source cell, reach a neighboring cell, and remain available there in a form capable of contributing to pharmacology. Linker class can affect the first of these steps through product properties, but does not by itself determine the outcome. Target-cell uptake, intracellular processing, product efflux, extracellular stability, diffusion, and recipient-cell sensitivity all matter.

One compact representation uses a source-cell product amount \(Y\) with first-order loss pathways: efflux at rate \(k_{\mathrm{out}}\) and all other intracellular loss at rate \(k_{\mathrm{other}}\). Under this model, the fraction of product leaving by efflux is \(k_{\mathrm{out}}/(k_{\mathrm{out}}+k_{\mathrm{other}})\). This is a simplifying competing-risks result: it assumes constant rates and eventual removal through one of the two pathways. The escaped amount is not identical to the amount entering a neighboring cell; extracellular loss and recipient-cell uptake would require additional states.

Assumptions, Diagnostics, and Alternative Comparisons

A mechanistic comparison is most interpretable when the shared delivery side is checked before attributing a difference to linker processing. Relevant checks include comparable DAR distributions and antibody concentrations, target binding and internalization, intracellular delivery, and linker stability in plasma. Then assess whether the proposed processing products and rates are supported by intracellular measurements. If uptake differs, an observed difference in product amount cannot be assigned to \(k_{\mathrm{cl}}\) versus \(k_{\mathrm{cat}}\) alone.

Useful diagnostics include the following:

• Check mass balance. In the simple model, the loss of \(Q\) must equal product formation before product loss is added. A predicted product amount exceeding payload equivalents delivered signals a unit, scaling, or state-accounting error.
• Inspect product identity. If measured catabolites do not match the assumed species, revise the product states before interpreting their permeability or potency.
• Separate formation from disappearance. A low intracellular product amount may reflect slow formation, rapid loss, or both. Time-course data are more informative than a single endpoint for distinguishing these explanations.
• Check linker stability outside cells. Premature plasma cleavage can create systemic free payload exposure and confound a model that assumes intracellular-only release.
• Test sensitivity to uncertain rates. If predicted product exposure or bystander ordering changes across plausible values of \(k_{\mathrm{rel}}\), product loss, or permeability, report that dependence rather than presenting one ordering as robust.

There are several reasonable levels of model detail. A matched mechanistic model represents ADC-derived material, processing, individual products, and product-specific loss or transport; it is useful when chemical identity and time courses are available. A lumped effective-release model combines processing and product availability into one empirical input; it can be practical when data do not identify the intermediate steps, but its parameters may not transfer to a different linker or cell type. A direct empirical comparison of matched-cell responses can establish an experimental difference under the tested conditions, but by itself does not identify whether that difference arose from cleavage rate, product potency, permeability, or another unmatched property.

The mechanistic model also has limits. First-order rates may not describe saturable processing or multiple lysosomal steps. A single source-cell compartment cannot resolve spatial gradients or heterogeneous intracellular trafficking. Product permeability and pharmacological activity may vary with cellular context. Consequently, predictions should be presented as conditional on the species, rates, and matching assumptions used, and tested against measurements that distinguish these mechanisms.

Common Pitfall

A common error is to conclude, “The cleavable ADC released more payload equivalents, so it must cause more killing and more bystander effect.” That reasoning jumps from formation to outcome. It ignores product loss and potency for cell-intrinsic effects, and it ignores membrane passage and extracellular survival for bystander effects. The correction is to compare the sequence of processes: delivered payload equivalents, processing, product-specific intracellular exposure, and, when relevant, escaped product and recipient exposure. Only then should response be interpreted using an appropriate pharmacology model.