Interpreting Bell-Shaped Dose Responses in Cell Engagers

Why a Cell Engager Can Become Less Effective at High Concentration

A bispecific cell engager must connect two binding partners—for example, an antigen on a target cell and CD3 on an effector cell—to support productive cell-cell bridging. Increasing engager concentration initially makes more bridging possible. At sufficiently high concentration, however, engager molecules can occupy the two target types separately: an engager bound to one target is less likely to encounter an unoccupied partner target for its other arm to bind. The measured response can therefore rise and then fall as concentration increases. This high-dose decline is commonly called a hook effect or prozone effect.

A bell-shaped curve alone does not establish that this mechanism is operating. Cell viability, target expression, assay timing, and the response measured can all affect curve shape. The useful question is whether the observed pattern is consistent with productive bridging at intermediate concentrations and excess-drug inhibition at high concentrations, under conditions where those interpretations can be checked.

A Reduced Model of Separate Arm Saturation

Let \(D\) be the free engager concentration, \(K_A\) the effective dissociation constant for binding target A, and \(K_B\) the effective dissociation constant for binding target B. In the reduced model below, each arm binds its target reversibly at equilibrium, and free engager concentration is approximately constant during the measurement. The arm-specific occupancy fractions are represented by the standard one-site expressions \(D/(K_A+D)\) and \(D/(K_B+D)\).

For a simple sequential picture, an engager first occupies target A; its other arm can then capture target B only when that partner is available rather than separately occupied by free engager. The fraction of A sites engaged is approximated by \(D/(K_A+D)\). The fraction of B sites remaining available is approximated by \(K_B/(K_B+D)\). Their product gives a dimensionless bridging proxy:

This proxy is not a universal exact law for a cell-cell synapse. It is a deliberately reduced equilibrium model that exposes the competing concentration effects: the first factor increases as A becomes occupied, while the second decreases as B becomes occupied separately. It assumes one effective site per target type, no material depletion of free engager, fixed target availability, and no substantial changes in binding or response during the measurement. Cell geometry, target density, multiple binding states, and downstream signaling can change the quantitative curve.

Why the Reduced Model Has a Peak

For positive \(D\), \(K_A\), and \(K_B\), the proxy is positive, so its maximum can be located by differentiating its logarithm. This avoids carrying the product rule through a long expression:

The denominator is positive. Consequently, the derivative is positive when \(D<\sqrt{K_AK_B}\), zero when \(D=\sqrt{K_AK_B}\), and negative when \(D>\sqrt{K_AK_B}\). Thus the proxy increases up to that concentration and decreases afterward. Its limiting behavior agrees with the interpretation: \(b(D)\) approaches zero as \(D\) approaches zero, and also approaches zero as \(D\) becomes very large.

This derivation gives a mechanistic reason for a bell-shaped bridging proxy, not a guarantee that every functional assay has its maximum at \(\sqrt{K_AK_B}\). A measured response can be a nonlinear, delayed, or noisy readout of bridging. The model’s peak is most useful as a qualitative and testable prediction under its assumptions.

Reading a Dose-Response Curve in Context

A dose-response curve should be interpreted as a sequence of observations, not just by its overall outline. A rise over the lower concentrations is consistent with increasing productive engagement. A reproducible decline at higher concentrations is consistent with a hook effect if high concentrations also make separate occupancy of both target types plausible and the decline is not explained by loss of viable cells or assay performance. In a functional assay, the y-axis might be target-cell lysis, activation, or cytokine release; those outcomes are not interchangeable and need not track bridging identically.

Occupancy can help assess whether the high-dose limb is mechanistically plausible. In the same hypothetical model, at \(D=30\) nM the separate occupancy fractions are \(30/(2+30)=0.9375\) for A and \(30/(8+30)\approx0.7895\) for B. The model’s estimated availability fraction for B is \(8/(8+30)\approx0.2105\). Thus high separate occupancy and reduced partner availability are plausible in this example. These calculations are model-based; they are not a substitute for measuring occupancy in the actual system.

Controls, Diagnostics, and Alternative Explanations

Interpret a bell-shaped curve as evidence for productive bridging followed by excess-drug inhibition only after checking whether the response depends on the intended two-target connection. Useful checks include confirming target expression on the relevant cells, testing target-negative cells or blocking one arm, and measuring an early bridging or activation readout alongside the later functional endpoint. Include viability and assay-performance controls at high concentrations, where nonspecific effects or measurement interference could mimic a decline. Replicates and a sufficiently broad concentration range are important: a few points around an apparent peak may not distinguish a real falling limb from variation.

• Target dependence: If removing or blocking either target leaves the curve essentially unchanged, the intended bridge is not a sufficient explanation for the response.
• High-dose controls: Check cell number, viability, and readout interference. A fall in signal caused by cell loss or assay saturation is not evidence of a hook mechanism.
• Concentration coverage: Sample both sides of the putative maximum. A monotonic rise over the tested range does not show that no hook effect exists; the range may not extend far enough.
• Endpoint and timing: Functional outputs can lag behind complex formation and may be affected by signaling or feedback. A functional maximum need not identify the exact concentration of maximum bridging.

The main alternatives answer different questions. A monotonic Emax model can describe a response that rises and plateaus, but it cannot represent a true high-dose decline. An empirical bell-shaped curve can summarize a rise and fall without claiming a mechanism. The reduced bridging proxy makes a specific mechanistic proposal—separate saturation of the two partners—but relies on stronger assumptions. A more detailed state-based binding model can represent additional states and constraints, but requires data sufficient to identify its parameters. Choose the simplest model that fits the observations and is appropriate to the mechanistic question.

Common Pitfall and Practical Limits

Wrong reasoning: “The response falls at the highest dose, so the engager must be causing the hook effect.” Correction: A falling limb is a pattern to explain, not a mechanism by itself. Check target dependence, viability, readout performance, and whether the concentration range and timing support the proposed saturation process. Conversely, a curve that has not yet fallen within the tested range cannot rule out a hook effect at higher concentrations.

The simple model assumes equilibrium, a stable free concentration, and fixed effective binding properties. Depletion of free engager, cell-cell contact geometry, target heterogeneity, internalization, or time-varying expression can change the curve or its peak. Downstream biology can further reshape the functional response. Therefore, the square-root peak is a result of the reduced model’s assumptions, not a dosing rule or a universal prediction for every cell engager.