Receptor pharmacology in a cell-based assay

Questions to investigate

1. Compare the potency of the two test agonists with the reference agonist. Which differences are supported by the confidence intervals of the paired comparison?
2. Which test agonist is a partial agonist? Lower the receptor expression: what happens to its maximum and its potency?
3. Run the Schild experiment with GN-404 and with GN-505. Which antagonist behaves competitively, and what shows this?
4. Switch off randomisation and switch on the edge effect. Why do responses exceed 100%, and why does Z′ not detect the problem?
5. Reduce the number of independent experiments to two, then raise it to eight. How do the confidence intervals change, and why do more technical replicates not have the same effect?
6. Raise the well-to-well variation until plates fail the Z′ criterion. What does the exclusion rule do to the analysis?

In a real laboratory

Cells stably expressing the receptor are seeded in 96- or 384-well plates the day before the assay and loaded with a fluorescent calcium indicator. Compounds are diluted in series, with the same DMSO concentration in every well; antagonists are added before the agonist and allowed to equilibrate. A plate reader records the fluorescence of each well before and after addition. Vehicle and maximal-agonist control wells on every plate give its Z′ factor (Zhang et al., 1999), and plate maps are randomised where possible. A biological replicate is an independent experiment on a separate cell passage and day; wells on one plate are technical replicates. Compounds can be coded so that the analyst is blind to them, and acceptance criteria and analysis are set out before the data are collected.

How the simulation works

Responses follow the operational model of agonism: each agonist has an affinity (KA) and an efficacy (τ) proportional to receptor expression, with a transducer slope of 1. GN-101 is the reference full agonist; GN-202, GN-303, GN-404 and GN-505 are hypothetical compounds, and which test agonist is partial and which antagonist non-competitive is left to be found (the true values can be shown with the results). A competitive antagonist increases the apparent KA by the factor 1 + [B]/KB. A non-competitive antagonist reduces τ by the same factor, so for an agonist with a large receptor reserve it first shifts the curve and then lowers its maximum, and at low concentrations its Schild plot can resemble that of a competitive antagonist. Experiments differ in signal window, basal fluorescence, receptor expression and affinities; plates drift slightly; wells vary in cell number, read noise and pipetting error, which accumulates down each dilution series; and the outer wells can show an edge effect. Antagonists are assumed to be at equilibrium. The time course of the calcium signal, dye loading, receptor desensitisation, compound solubility and binding to plastic are not simulated, and the values are approximate.