Diagnosing Receptor-Theory Models with Binding and Response Data

Why Combine Binding and Functional Data?

A binding assay describes how a ligand associates with receptors; a functional assay describes the response produced by that ligand. The two measurements are related, but they are not interchangeable. A ligand can occupy a substantial fraction of receptors while producing a smaller response, or produce a near-maximal response without occupying every receptor. Receptor reserve and response amplification can make the functional curve appear more potent than receptor binding alone would suggest.

The diagnostic task is to ask whether one model can account for both kinds of data under compatible conditions. A mismatch does not, by itself, prove a particular mechanism is wrong: the assays may differ in receptor state, temperature, incubation time, or measurement scale. It does signal that affinity, efficacy, transduction, and experimental conditions should not be conflated.

Definition: For reversible binding to one class of independent sites at equilibrium, fractional occupancy at free agonist concentration \(C\) is \(\theta=C/(K_D+C)\), where \(K_D\) is the equilibrium dissociation constant. In a functional model, \(K_A\) denotes the agonist affinity parameter used by that model; it may agree with a binding \(K_D\) measured under matched conditions, but equality is an empirical assumption to check.

Models and Their Assumptions

The occupancy-proportional model assumes that response above baseline is directly proportional to occupancy. If \(E_0\) is baseline and \(E_{\mathrm{sys}}\) is the response reached at full occupancy, its one-site form is:

E(C)=E_0+E_{\mathrm{sys}}\frac{C}{K_D+C}.

This model assumes equilibrium, one relevant binding-site population, negligible ligand depletion, and a linear relation between occupied receptors and response. Its functional half-maximal concentration is \(K_D\), and its plateau above baseline is \(E_{\mathrm{sys}}\). Thus, a functional curve with a lower half-maximal concentration or a plateau that cannot represent the system maximum challenges at least one of those assumptions.

The operational model of agonism allows receptor activation to be amplified and allows an agonist to have limited efficacy. In the common form with transducer slope \(n=1\), it is:

E(C)=E_0+E_{\mathrm{sys}}\frac{\tau C}{K_A+(1+\tau)C}.

Here \(\tau\) is an operational efficacy parameter: larger \(\tau\) means that a given receptor occupancy generates more stimulus relative to the system’s transduction scale. \(E_{\mathrm{sys}}\) is the system maximum above baseline, not necessarily the maximum observed for a particular agonist. The model presumes a stable baseline and system maximum, equilibrium binding represented by \(K_A\), and a specified transduction form. It is a structured description, not proof that a unique molecular mechanism has been identified.

Algebra gives two useful relationships. The observed maximum above baseline is \(E_{\mathrm{sys}}\tau/(1+\tau)\), and the functional concentration producing half of that agonist’s own maximum is \(K_A/(1+\tau)\). Therefore, operational amplification can lower functional EC50 relative to \(K_A\) and can make the observed plateau lower than \(E_{\mathrm{sys}}\), especially for a partial agonist.

Deriving the Operational Model Diagnostics

For \(n=1\), divide the operational response above baseline by its plateau. The plateau is \(E_{\mathrm{sys}}\tau/(1+\tau)\), so:

\frac{E(C)-E_0}{E_{\mathrm{sys}}\tau/(1+\tau)} = \frac{\tau C}{K_A+(1+\tau)C}\frac{1+\tau}{\tau} = \frac{C}{K_A/(1+\tau)+C}.

The final expression is one-half when \(C=K_A/(1+\tau)\). Also, as \(C\) tends to infinity, the response above baseline tends to \(E_{\mathrm{sys}}\tau/(1+\tau)\). These results justify checking whether the measured plateau and EC50 imply plausible operational parameters rather than interpreting EC50 as a direct affinity estimate.

When the system maximum is known or defensibly fixed, and the observed agonist maximum \(E_{\max}\) is below it, rearranging the plateau expression yields:

\tau=\frac{E_{\max}}{E_{\mathrm{sys}}-E_{\max}}, \qquad K_A=EC_{50,\mathrm{functional}}(1+\tau).

These calculations are diagnostics, not substitutes for fitting and uncertainty assessment. In particular, if \(E_{\max}\) is close to \(E_{\mathrm{sys}}\), the denominator is small and inferred \(\tau\) can become very large and sensitive to small measurement errors.

Worked Example: Binding and an Amplified Agonist Response

Suppose a hypothetical binding assay gives \(K_D=10\) nM. Its measured occupancy is 0.50 at 10 nM and 0.75 at 30 nM, as predicted by \(C/(K_D+C)\). In a compatible functional assay, the baseline-subtracted response as a percentage of system maximum is 26.7% at 1 nM, 48.0% at 3 nM, 66.7% at 10 nM, 75.0% at 30 nM, and approaches 80% at high concentrations. Compare the one-site occupancy and operational interpretations.

Under occupancy proportionality, with \(E_{\mathrm{sys}}=100\%\) and \(K_D=10\) nM, the predicted response at 10 nM is \(100\%\times10/(10+10)=50.0\%\). That is below the hypothetical observed 66.7%. This single difference is not decisive, but the full curve also plateaus near 80%, rather than 100%, so direct proportionality to occupancy with a 100% full-occupancy response does not describe these data.

For the operational model, take \(K_A=10\) nM, \(\tau=4\), and \(E_{\mathrm{sys}}=100\%\). At 10 nM:

E(10)-E_0 = 100\%\times\frac{4(10)}{10+(1+4)(10)} = 100\%\times\frac{40}{60} = 66.7\%.

At 3 nM the prediction is \(100\%\times12/(10+15)=48.0\%\), and at 30 nM it is \(100\%\times120/(10+150)=75.0\%\). The predicted plateau is \(100\%\times4/5=80.0\%\), and the functional EC50 is \(10/5=2.00\) nM. The operational model therefore accounts for the hypothetical curve while retaining an affinity parameter matching the binding \(K_D\). This agreement is plausible under matched assay conditions; it does not independently establish that the two parameters must always be equal.

Testing Competitive Antagonism Against Binding

A reversible competitive antagonist competes with agonist for the same receptor population, is assumed to bind reversibly at equilibrium, and does not itself reduce the attainable system maximum. Let \(B\) be free antagonist concentration and \(K_B\) its equilibrium affinity constant. Under simple competition, the agonist concentration required for a given receptor occupancy is multiplied by \(1+B/K_B\). The dose ratio is therefore:

DR=\frac{EC_{50,\mathrm{with\ antagonist}}}{EC_{50,\mathrm{control}}} =1+\frac{B}{K_B}.

For the operational model with this competitive antagonist, the corresponding \(n=1\) response is:

E(C,B)=E_0+E_{\mathrm{sys}} \frac{\tau C}{K_A(1+B/K_B)+(1+\tau)C}.

This predicts a rightward concentration-response shift and the same limiting agonist maximum, provided the antagonist remains surmountable and the model assumptions hold. A binding estimate of \(K_B\) can be checked against functional shifts. If shifts are not parallel, maxima change, or estimated affinity varies systematically with antagonist concentration, simple competitive antagonism may be inadequate. Insurmountable and irreversible mechanisms are distinct possibilities, not conclusions to draw from one curve alone.

Worked Example: Checking an Antagonist-Induced Shift

Use the operational agonist in the previous example, with \(K_A=10\) nM and \(\tau=4\). Suppose a hypothetical binding experiment estimates \(K_B=5\) nM for a competitive antagonist, and the functional assay uses \(B=15\) nM. Calculate the expected shift and check the response at an agonist concentration of 8 nM.

The predicted dose ratio is \(1+15/5=4.00\). Without antagonist, the operational functional EC50 is \(10/(1+4)=2.00\) nM, so the expected EC50 with antagonist is \(4.00\times2.00=8.00\) nM. Equivalently, the antagonist-adjusted affinity term is \(10(1+15/5)=40\) nM.

At \(C=8\) nM, the predicted baseline-subtracted response is:

100\%\times\frac{4(8)}{40+(1+4)(8)} = 100\%\times\frac{32}{80} = 40.0\%.

Without antagonist, the response at \(C=2\) nM is \(100\%\times8/[10+5(2)]=40.0\%\). The matched responses are consistent with a fourfold rightward shift, and the operational model predicts the same 80.0% limiting response. If the observed antagonist condition instead had a reliably lower plateau, this simple competitive model would fail an important qualitative check even if its estimated EC50 shift were near fourfold.

Affinity Consistency and Model Failure

A useful cross-check is whether binding and functional estimates can be reconciled without implausible assumptions. For example, a functional EC50 below binding \(K_D\) is not automatically contradictory: operational amplification can produce that pattern. The relevant question is whether a plausible \(\tau\), system maximum, and assay interpretation account for both the shift in potency and the observed maximum.

Worked Example: Quantifying an Affinity Mismatch

Suppose hypothetical binding data give \(K_D=10\) nM. Functional data, on a scale where the system maximum is \(E_{\mathrm{sys}}=100\%\), are adequately summarized by an observed agonist maximum of \(90\%\) and a functional EC50 of \(0.500\) nM. Under the \(n=1\) operational model, calculate the implied efficacy and affinity, then interpret their relationship to binding.

The implied efficacy is \(\tau=90/(100-90)=9.00\). The implied operational affinity is \(K_A=0.500(1+9.00)=5.00\) nM. Thus the functional fit implies an affinity parameter twofold lower than the binding \(K_D\) of 10 nM. The discrepancy is not explained merely by saying “there is receptor reserve”: the operational calculation already accounts for amplification, yet the inferred \(K_A\) still differs from the binding estimate.

That difference could reflect uncertainty, unmatched assay conditions, a binding model that is too simple, or a functional model that is inadequate. It could also reflect a genuine difference between the fitted affinity parameter and the affinity measured in the binding assay. The numbers alone do not decide among those explanations. The correct diagnostic is to inspect uncertainty, assay comparability, and residual patterns before asserting a mechanism.

Practical Diagnostics and Limitations

Model checks should consider the entire concentration range, not only a fitted EC50. Plot observed binding and response against concentration on a logarithmic concentration axis, overlay model predictions, and inspect residuals against concentration. A repeated S-shaped residual pattern, a systematic mismatch at low concentrations, or an apparent plateau that changes with assay conditions can indicate a wrong curve form, an incorrect baseline, inadequate equilibration, or an unmodeled receptor population. The earlier material on heterogeneous receptor populations is relevant when a single smooth transition fails to describe the data.

Check whether concentrations represent free ligand at the receptor. Ligand depletion, nonspecific binding, incomplete equilibration, or errors in concentration units can make apparent affinity misleading. Verify that binding and functional measurements concern comparable receptor populations and conditions before demanding exact equality of \(K_D\) and \(K_A\). Also check whether the functional maximum is independently known: if both \(E_{\mathrm{sys}}\) and \(\tau\) are estimated from a limited concentration range, their values may be weakly determined.

The main alternatives answer different questions. The occupancy-proportional model is parsimonious when response tracks occupancy directly and full occupancy corresponds to the measured system maximum. The operational model is appropriate when efficacy and amplification are needed to explain functional potency or a submaximal agonist plateau. A competitive-antagonist model adds a binding-based prediction for concentration-response shifts. More flexible slopes or multiple receptor populations may improve fit, but extra parameters should be justified by data and biology; a better numerical fit alone does not establish a mechanism.

No model can repair incompatible measurements automatically. A fitted affinity may be precise within a model and still be biologically misleading if equilibrium, free concentration, receptor homogeneity, or response-scale assumptions fail. Treat binding-response agreement as a consistency check with explicit assumptions, not as a universal identity.

Common pitfall: “The functional EC50 is below the binding \(K_D\), so the data are impossible.” This overlooks operational amplification: for \(n=1\), \(EC_{50,\mathrm{functional}}=K_A/(1+\tau)\). The correction is to compare the functional EC50 and maximum jointly, infer or fit operational parameters under stated assumptions, and then assess whether the resulting \(K_A\) is consistent with binding. A remaining mismatch warrants investigation, not an automatic mechanistic claim.

Check Your Understanding

Answer each question using the stated model assumptions. Brief solutions follow the questions.

For an operational agonist with \(K_A=12\) nM, \(\tau=3\), and \(E_{\mathrm{sys}}=100\%\), find the functional EC50 and the observed maximum above baseline.
A competitive antagonist has \(K_B=4\) nM and is tested at \(B=12\) nM. What dose ratio does the model predict?
A one-site occupancy-proportional model predicts 50% of system maximum at \(C=K_D\). State one assumption whose failure could make that functional prediction inappropriate.
Binding suggests \(K_D=8\) nM, while functional data imply \(K_A=8\) nM after operational amplification is accounted for. What can be concluded, and what cannot be concluded, from this agreement?

Solutions. (1) \(EC_{50}=12/(1+3)=3.00\) nM; the maximum is \(100\%\times3/4=75.0\%\) above baseline. (2) \(DR=1+12/4=4.00\). (3) For example, response may not be proportional to occupancy because receptor reserve or nonlinear transduction is present; the one-site equilibrium, negligible-depletion, and correct system-maximum assumptions also matter. (4) The estimates are consistent under the fitted models and compatible conditions; agreement does not prove a unique mechanism, exact equality in other tissues, or that either model’s assumptions are true.