Proving Allosteric Modulation in Peptide Receptor Assays

Calling a compound an allosteric modulator is a claim about topography, and it carries obligations that a potency number cannot discharge. The claim is that the compound occupies a site distinct from the one the endogenous ligand uses, that both sites can be occupied at once, and that the resulting ternary complex has properties neither binary complex has. G protein-coupled receptors are natural allosteric proteins in exactly this sense, because agonist-mediated signalling requires a conformational change transmitted between two topographically distinct binding sites, one for the agonist and another for the G protein ^[1]. A modulator adds a third site to that circuit.

The reference case for the assays that follow is gallamine at muscarinic receptors, which binds a site distinct from the conventional muscarinic ligand binding site and modulates the binding of both agonists and antagonists at that conventional site; its effects on muscarinic ligand binding were much greater in heart than in other tissues ^[2]. The tissue dependence in that early work is worth holding onto, because it is the first hint that what you measure is a property of the whole preparation, not of the compound.

The binding signature: affinity moves, capacity does not

At the level of equilibrium binding the theory is specific. An allosteric ligand changes the affinity of a ligand bound at the primary recognition site without affecting that ligand's binding capacity, and the magnitude of the change depends on the amount of cooperativity, positive or negative, between the two sites ^[3]. In a saturation experiment run in the absence and presence of a fixed modulator concentration, that predicts a shifted K_d and an unchanged B_max. A result in which B_max falls is telling you something else — receptor loss, a pseudo-irreversible occupant, or a preparation problem — and should be chased before any cooperativity constant is fitted.

Cooperativity and the ceiling

The cooperativity factor, conventionally α, scales the affinity of the orthosteric ligand when the allosteric site is occupied: α > 1 for positive cooperativity, α < 1 for negative, α = 1 for neutral. Because the modulator can only act through the ternary complex, its effect is exhausted once the allosteric site is saturated. This produces a ceiling level to the allosteric effect, which is one of the properties that distinguishes allosteric modulators from classic orthosteric ligands ^[1].

That ceiling is what an equilibrium binding experiment can test directly. Titrate the test compound against a fixed radioligand concentration and follow specific binding. A competitor at the orthosteric site keeps displacing until specific binding approaches zero. A negative modulator drives specific binding down to a plateau set by α and stops there, however much more you add. The same saturability yields the modulator's own affinity: the dissociation constant of the allosteric ligand can be calculated from the concentration producing half of its maximal effect on radioligand binding ^[3].

Two things have to be true before that plateau counts as evidence. It has to be established rather than inferred — several concentrations above the inflection point, with the top concentration constrained by solubility rather than by convenience, because an apparent plateau caused by running out of soluble compound looks identical on the plot. And the plateau describes a pair. Cooperativity is a property of the modulator together with the ligand you happened to label, not of the modulator alone.

Efficacy modulation, and the operational model

An affinity cooperativity constant does not exhaust the mechanism. The earlier mathematical models of allosteric interaction concentrated on the modulator's effect on orthosteric ligand binding and ignored possible effects on the ability of orthosteric ligands to cause receptor activation; an extension built on the two-state model of receptor activation explicitly includes the latter, and its predictions generate a set of criteria that should be tested before an allosteric modulator's effects can be quantified in a manner that is not system-dependent ^[6]. In practice this means a modulator has at least two independent handles: cooperativity on binding and a separate scaling of the orthosteric agonist's efficacy.

The operational model of allosterism folds both into a single fit, with transducer terms carrying the system. It has been validated at mGlu₅, where a large set of chemically diverse modulators allowed quantitative estimation of affinity and cooperativity values; affinity estimates derived from functional assays fitted well with affinities measured in radioligand binding experiments for both positive and negative allosteric modulators, and the model could be applied to modulator potency curves in combination with maximal fold-shift data to derive estimates of modulator affinity ^[8]. The convergence of functional and binding estimates is the point: it is the internal check that the model is describing the compound rather than the assay.

Probe dependence is the diagnostic and the trap

Because cooperativity belongs to a pair, one modulator can have several signs at one receptor. Strychnine at muscarinic receptors is the clearest early demonstration. It binds the unliganded m1, m2 and m4 receptors with an affinity of approximately 10⁵ M^-1 and is 5–10-fold weaker at m3; it is positively cooperative with N-methylscopolamine at m2 and m4, neutral at m1 and negative at m3; and with acetylcholine it is negatively cooperative, though the degree of cooperativity is relatively low, 2–7-fold, and particularly low at m1 and m4 ^[4]. Read that as an instruction rather than a curiosity: at m2 and m4 the same compound is a positive modulator of a labelled antagonist and a negative modulator of the endogenous agonist.

The consequence for selectivity work is sharper still. LY2033298 was described as a modulator that preferentially potentiates acetylcholine at M₄ receptors, but it also binds M₂ and mediates robust positive or negative allosteric effects there depending on which orthosteric ligand is used as the probe. Its effects on binding at M₂ were positive in every instance tested, whereas its effects on signalling were positive or strongly negative depending on both the agonist and the pathway read, [³⁵S]GTPγS or ERK1/2 phosphorylation. The apparent subtype selectivity therefore did not arise from selective affinity for a poorly conserved allosteric site but from subtype-selective cooperativity with acetylcholine at a common allosteric site ^[9]. A selectivity panel that measures only modulator affinity across subtypes will rank a compound like this wrongly.

For peptide receptors the same effect appears across the family of endogenous agonists. At the GLP-1 receptor, two allosteric modulators were tested against the naturally occurring peptide agonists and a peptide mimetic. One selectively augmented cAMP signalling but did so in an agonist-dependent way, with the greatest effect on oxyntomodulin, a weaker effect on truncated GLP-1 peptides and negligible effect on the other peptide responses, and those effects were driven principally by parallel changes in peptide agonist affinity. The other selectively modulated calcium signalling, and only for the truncated GLP-1 peptides and exendin-4, not for oxyntomodulin or the full-length peptides ^[10]. If you screen a peptide receptor against one surrogate agonist in one pathway, that is precisely and only what you have characterised.

Neutral cooperativity is a genuine result, not a failed run. A β-arrestin recruitment screen at the μ-opioid receptor returned both positive allosteric modulators and silent allosteric modulators ^[11] — compounds that occupy the allosteric site without altering the response to the probe in use.

Radioligand dissociation kinetics

The equilibrium experiments above are suggestive; the dissociation experiment is the one a competitor cannot pass. A ligand restricted to the vacant orthosteric site has no route to influence the departure of a ligand already bound there, so allosteric modulation of a receptor may be proved experimentally by demonstrating an altered radioligand dissociation in the presence of the modulator ^[5].

The design is straightforward. Equilibrate the receptor preparation with radioligand, then initiate dissociation either by large-volume dilution or by adding a saturating concentration of unlabelled orthosteric ligand, and follow the time course with and without the test compound. Both initiation methods have to give the same control k_off; if adding the unlabelled orthosteric ligand accelerates dissociation relative to dilution, you have an allosteric effect of the orthosteric ligand itself, or incomplete dilution allowing rebinding. Two-point kinetic experiments provide a screening-type approach to quantifying the delay in dissociation, but the analysis carries a pitfall that can lead to a suboptimum determination of allosteric potency when dissociation is monophasic ^[5].

Be careful what a null licenses. Unchanged dissociation shows neutral cooperativity with that radioligand under those conditions; it does not exclude an allosteric site, as the existence of silent modulators makes concrete ^[11], and it does not predict the sign of cooperativity with a different probe ^[4]. Conversely a positive result establishes co-occupancy but does not by itself give you α, which needs the equilibrium work. Because kinetic and equilibrium assays interrogate different aspects of the same interaction, agreement between them is the verification step. In the strychnine study, a semiquantitative affinity-ratio protocol, detailed binding assays, guanosine-5'-O-(3-[³⁵S]thio)triphosphate functional assays and two non-equilibrium binding assays gave quantitatively similar and internally consistent estimates of the cooperativity parameters ^[4]. More generally, detection and quantification of allosteric effects usually relies on a combination of equilibrium binding, non-equilibrium kinetic and functional signalling assays ^[1].

Receptor reserve, allosteric agonism, and the fold-shift you cannot interpret

The evidence most often offered for modulation is a leftward shift of an agonist concentration-response curve, and on its own it is the least diagnostic of the experiments described here, because the size of that shift is partly a property of the preparation. If the agonist has a large receptor reserve, or if the allosteric ligand influences only affinity, the modulator's dissociation constant can be recovered from the EC₅₀ shift; but if the agonist is a partial agonist with little receptor reserve, the shift in EC₅₀ caused by the allosteric ligand may depend on the affinity of the agonist ^[3]. A fold-shift is therefore not a system-independent quantity, which is why criteria have to be satisfied before modulator effects are quantified in a system-independent way ^[6].

The pathway read matters on top of that. Modulation bias was observed for positive modulators at mGlu₅ when Ca²⁺ mobilisation and ERK1/2 phosphorylation data were compared ^[8]. A fold-shift reported without naming the pathway, the probe agonist and the expression system is not a number anyone can reuse.

The commonest artefact is the modulator's own activity. Allosteric ligands can alter the efficacy of orthosteric ligands and can also activate seven-transmembrane receptors in their own right ^[12]. A compound with intrinsic efficacy raises the baseline of an agonist curve and inflates apparent potentiation, so run the modulator alone across its full concentration range in the same assay and on the same plates before interpreting any shift. Anchor the mechanism with mutagenesis where the receptor allows it: at M₂, residues Tyr177 and Trp99^3.28 contributed to LY2033298 binding, while the orthosteric-site residues Tyr104^3.33 and Tyr403^6.51 contributed to the ligand's ability to impose pathway-biased modulation ^[9]. That separation — binding determinants in one place, transmission determinants in another — is what a mechanistic claim looks like when it is supported.

Peptides as allosteric ligands, and the surfaces they can reach

A peptide test article is large, polar and confined to the aqueous phase, which restricts it to solvent-exposed epitopes. The extracellular vestibule is the obvious one. The structure of the M₂ receptor bound simultaneously to the orthosteric agonist iperoxo and the positive allosteric modulator LY2119620 shows the modulator recognising a largely pre-formed binding site in the extracellular vestibule of the agonist-bound receptor, inducing a slight contraction of that outer pocket ^[13]. A largely pre-formed site in the extracellular vestibule, rather than a buried pocket, is the kind of surface a peptide can engage.

Protease-activated receptor 2 illustrates the range of accessibility on a single peptide-ligand receptor, whose tethered peptide ligand is unmasked by proteolytic cleavage of the N terminus. One antagonist binds a fully occluded pocket near the extracellular surface and shows slow binding kinetics; a second binds a remote allosteric site outside the helical bundle; and a blocking antibody antigen-binding fragment binds the extracellular surface and prevents access of the tethered ligand to the peptide-binding site ^[14]. An extrahelical site outside the bundle is not addressable by a peptide arriving from solution. The extracellular face, as the antibody fragment shows, is.

Endogenous peptides already work this way. 5-HT-moduline, a tetrapeptide, interacts specifically with 5-HT_1B/1D receptors by a noncompetitive mechanism; [³H]5-HT-moduline binds a single population of sites with K_d 0.4 nM in rat and 0.8 nM in guinea pig brain, and 0.2 nM and 0.6 nM in cells transfected with the 5-HT_1B and 5-HT_1Dβ receptors respectively, with binding specific to the LSAL sequence; functionally the peptide decreased the inhibitory effect of a 5-HT_1B agonist on evoked release of [³H]5-HT from a synaptosomal preparation ^[15]. Note the analyte: a four-residue peptide, radiolabelled and used as its own probe, which is what makes the site assignment tractable.

Class C receptors present a far larger extracellular target. γ-Glutamyl di- and tripeptides, including glutathione and its S-methyl and S-propyl analogues and the dipeptides γ-Glu-Ala and γ-Glu-Cys, are positive allosteric modulators of the calcium-sensing receptor; γ-glutamyl peptides were identified previously as positive allosteric modulators of Ca²⁺_o-dependent intracellular Ca²⁺ mobilisation in HEK-293 cells that bind in the Venus flytrap domain, and in HEK-293 cells expressing the receptor a T145A/S170T double mutant showed markedly impaired Ca²⁺ mobilisation and cAMP suppression responses to S-methylglutathione as well as to L-Phe, indicating that the peptides and the L-amino acids act through a common mechanism ^[16]. The mutant is doing the work there, converting a correlation into a site assignment.

What an unmodified peptide cannot do is reach the intracellular face. Peptides derived from the third intracellular loop of protease-activated receptors 1 and 2 and of the melanocortin-4 receptor became agonists or antagonists of receptor–G protein signalling on attachment of a palmitate group; these lipidated peptides required the presence of their cognate receptor for activity, were highly selective for receptor type, and by mutational analysis did not activate G proteins by the same mechanism as the intact third intracellular loop but instead required the receptor C-tail ^[17]. The lipid anchor is a delivery solution, and it changes the mechanism rather than merely relocating the peptide.

How this sits against the neighbouring methods

The radioligand binding workflow supplies the substrate for all of this: a saturation experiment fixing K_d and B_max for the labelled probe, and a competition format into which the modulator titration is dropped. The dose-response workflow supplies the fold-shift, which this guide argues should never travel without its probe, pathway and expression system attached. And the selectivity workflow needs restating for allosteric compounds: because selectivity can live in cooperativity rather than in affinity ^[9], a subtype panel has to be run as a modulation experiment against a defined orthosteric probe at each subtype, not as a binding panel. The prospect that allosteric ligands can engender stimulus bias in orthosteric ligand signalling ^[7] is the reason the probe and the pathway belong in the protocol, chosen before the data rather than after.