Bystander Killing From Membrane-Permeable Payload

From Payload Release to Neighboring-Cell Killing

An antibody-drug conjugate (ADC) can be taken up by an antigen-positive cell and release payload intracellularly. If released payload leaves that cell, enters the surrounding tissue, and crosses the membrane of a nearby antigen-negative cell, it can expose that neighbor to a cytotoxic drug. This is bystander killing: killing of cells that need not themselves bind or internalize the ADC. It depends on a chain of processes, not on membrane permeability alone. Release, extracellular transport, local loss, membrane entry, and the cell’s response all matter.

The preceding material on intracellular payload release supplies the release flux that drives this model. Here, treat that flux as an input; the aim is to follow released payload into neighboring cells and compare predictions when membrane entry is efficient or limited. The equations below describe a deliberately reduced local tissue model. Concentrations and rates are hypothetical, and the model should be expanded if the data resolve spatial gradients, important cell uptake, or time-varying release in more detail.

A simple extracellular transport block is

This is a lumped representation: \(b\) is the diagonal first-order disappearance coefficient, while \(g\) represents coupling between the two local pools. For a stable transport block, \(b>g\geq 0\). The equations allow material in the neighborhood to contribute back to the source-region pool. They do not resolve a tissue geometry or a concentration gradient within either region. The intracellular response is represented by

This last equation treats the extracellular neighborhood concentration as the driving concentration for entry. If cell uptake substantially depletes that pool, the uptake sink must also be included in the extracellular mass balance; treating it only as a driver is then an approximation, not a complete mass-conserving model.

Steady Transport and Its Reduction

At steady state, the neighborhood equation gives \(C_n=(g/b)C_s\). Substitution into the source equation yields a reduced balance for the source pool. The reduction is useful because it shows how transport-mediated return changes the effective coefficient relating release to source concentration.

The denominator is positive when \(b>g\): \(b-g^2/b=(b^2-g^2)/b>0\). The stability condition also follows from the transport matrix, whose eigenvalues are \(-b+g\) and \(-b-g\); both are negative if \(b>g\geq0\). The reduction assumes the extracellular pools have reached steady state. It should not be used as a description of a short release pulse before equilibration.

For constant \(C_n\) and \(k_x>0\), the intracellular steady state is \(X_n=(k_{\mathrm{in}}/k_x)C_n\). Thus, for the same local extracellular concentration, greater membrane entry increases intracellular exposure. For a time-varying concentration, the full differential equation matters: accumulation lags behind \(C_n\), and intracellular loss can continue after the extracellular concentration falls.

Why Permeability and Spatial Transport Must Be Separated

The parameter \(g\) describes coupling between extracellular regions; \(k_{\mathrm{in}}\) describes entry across a neighboring cell’s membrane. They are not interchangeable. A payload can diffuse through tissue but enter cell membranes poorly, producing a nonzero \(C_n\) and low \(X_n\). Conversely, good membrane entry cannot create substantial bystander exposure if payload release is small or the extracellular payload is lost before reaching neighboring cells.

For a time-varying neighborhood concentration, the intracellular equation has the solution, for \(t\geq0\),

This expression follows by multiplying \(dX_n/dt+k_xX_n=k_{\mathrm{in}}C_n\) by the integrating factor \(e^{k_xt}\), integrating from 0 to \(t\), and dividing by \(e^{k_xt}\). It shows that intracellular concentration is a loss-weighted history of extracellular exposure, not simply the current \(C_n\). When \(C_n\) is constant and \(X_n(0)=0\), it reduces to \(X_n(t)=(k_{\mathrm{in}}C_n/k_x)(1-e^{-k_xt})\).

Checking Assumptions and Comparing Approaches

A steady compartment model is most informative when release and transport are approximately linear over the modeled concentration range, the parameters are reasonably constant, and each compartment can be treated as well mixed. The simple killing calculation additionally assumes a specified hazard function, constant hazard during the exposure, and no cell proliferation or cell-to-cell variability. The hazard formula is an illustrative pharmacodynamic closure; a real application needs a response relationship appropriate to its payload, cell type, and exposure schedule.

Useful diagnostics include checking whether predicted concentrations are nonnegative, whether the transport rates imply a stable system, and whether predicted concentration time courses agree with measured extracellular or intracellular payload. In the steady two-pool system, \(b>g\) ensures positive stable decay rates. A fitted \(g\geq b\) is a warning that the model or parameter interpretation is inadequate, not evidence for an indefinitely accumulating payload. Compare predicted and observed time courses as well as endpoint viability: similar endpoint killing can arise from distinct exposure histories.

The main alternatives answer different questions. A spatial reaction-diffusion model is preferable when gradients across several cell diameters or tissue barriers are important; it represents transport in space rather than compressing it into \(g\). A multicompartment model can represent more than one neighborhood and may be sufficient when only a few regions are distinguishable. A direct empirical relationship between release and viability can be useful for prediction in the measured system, but it cannot distinguish poor membrane entry from poor tissue transport without additional measurements. Mechanistic detail should be added only when the available observations can inform it.

A frequent mistake is to interpret a larger permeability parameter as proof of greater bystander killing in every setting. The correction is that membrane entry is only one link in the chain: the neighboring extracellular concentration must also be sufficient, intracellular payload must persist long enough, and the response relationship must convert exposure into injury. Another pitfall is to equate antigen-negative cells with unexposed cells. They may lack ADC-mediated uptake yet still receive released payload from nearby antigen-positive cells.

Limitations and Model Use

A bystander model is a mechanistic hypothesis, not a guarantee that a permeable payload will improve therapeutic index. Permeation may increase exposure of antigen-negative tumor cells, but it may also expose nearby normal cells. The reduced equations do not represent tissue architecture, payload binding to extracellular components, variable cell density, intracellular sequestration, or heterogeneous sensitivity. If these processes affect the question of interest, add them explicitly and seek data that can identify the new parameters.

The steady-state reduction is especially fragile when release is pulsed or changes rapidly. In that setting, solve the differential equations over time and use the exposure history in the response model. Also test whether the model’s aggregate extracellular loss remains plausible when permeability or cell density changes. If uptake is a major sink, it must be coupled to the extracellular balance rather than treated solely as an intracellular input. These checks prevent a convenient reduced model from being mistaken for a complete mass balance.