What the GLP-1 receptor actually does when exenatide binds it

Ask what a GLP-1 receptor agonist does and you will usually be told that it makes you less hungry. That is a consequence, several steps downstream, of something considerably more specific: a peptide occupying an orthosteric binding site on a class B G-protein-coupled receptor, stabilising a conformation that couples preferentially to Gα_s, raising intracellular cyclic AMP, and — depending on the ligand — recruiting beta-arrestin to a greater or lesser degree. Every clinically interesting property of this drug class, including the ones patients notice first, is a consequence of how a particular molecule performs that sequence.

A class B receptor, and why that matters

The GLP-1 receptor belongs to class B of the G-protein-coupled receptor superfamily — the secretin-like receptors — which is a structural classification with practical consequences. Class B receptors have a large extracellular domain that captures the C-terminal portion of a peptide ligand first, in what is usually described as a two-domain binding model: the extracellular domain provides affinity, and the N-terminal residues of the peptide then insert into the transmembrane bundle to provide activation.

That architecture is why these receptors are difficult small-molecule targets and why, for two decades, every marketed agonist was a peptide. It is also why the orally available non-peptide agonists now in late-stage development are genuinely notable pharmacology rather than a formulation trick: they bind a site that a peptide does not occupy in the same way, and they activate the receptor through a partially distinct mechanism.¹

The consequence for a reader trying to compare molecules is that structural class predicts a great deal about route, durability and formulation, and rather less about efficacy.

Biased agonism, stated carefully

When the GLP-1 receptor is activated it can couple to Gα_s, raising cyclic AMP, and it can recruit beta-arrestin, which contributes to receptor internalisation and desensitisation. An agonist that favours the first over the second is described as G-protein-biased. The therapeutic argument for bias is that sustained cAMP signalling without proportionate internalisation should produce a more durable effect at the same occupancy.

The evidence for that argument is real but narrower than its popularity suggests. Bias is measured in transfected cell systems at receptor densities that bear no relationship to a beta cell or a vagal afferent, and the translation from a bias factor in vitro to a clinical difference in vivo has been demonstrated convincingly for very few ligands.² The Journal’s position is that bias is a legitimate and probably important variable, that it is one of several plausible explanations for the differences observed between molecules, and that anybody presenting it as the explanation is ahead of the data.

Selectivity, potency and efficacy are three measurements

Three quantities are routinely conflated in discussions of this class. Affinity is how tightly a ligand binds, usually reported as a dissociation constant. Potency is the concentration producing half-maximal response, reported as an EC₅₀. Efficacy is the maximal response achievable, reported relative to a reference agonist. A molecule can be more potent and less efficacious than another, and a molecule can bind a second receptor with high affinity and produce almost no response there.

Selectivity is the ratio of activities across receptors, and it is where the current pipeline diverges most sharply. Reported GIP-to-GLP-1 activity ratios for dual agonists vary by more than an order of magnitude between molecules; glucagon receptor arms in triple agonists vary similarly. Those ratios are properties of the sequence and they are not adjustable by dose. Two molecules with different ratios are different drugs at every dose, which is the reason head-to-head trials cannot be replaced by cross-trial comparison.³

How you get seven days out of a two-minute peptide

Three engineering strategies account for essentially every long-acting agonist on the market. The first is substitution at the DPP-4 cleavage site: replacing the alanine at position 8 with a residue the enzyme cannot process removes the fastest route of degradation. The second is acylation with a fatty-acid chain, which promotes reversible binding to serum albumin; albumin-bound drug is protected from renal filtration and enzymatic attack, and dissociates slowly to provide a circulating depot. The third is fusion to a large carrier — an immunoglobulin Fc fragment, for instance — which raises the hydrodynamic radius above the glomerular filtration threshold.

Semaglutide uses the first two, with a C18 diacid linked through a spacer. Liraglutide uses a shorter C16 chain and achieves roughly thirteen hours rather than seven days, which is a useful demonstration of how much the chain contributes. Dulaglutide takes the fusion route. The strategies are not interchangeable and they produce different distribution and clearance behaviour, not merely different durations.⁴

The oral non-peptide agonists

An orally bioavailable small molecule that activates a class B GPCR was, for a long time, considered close to impossible. The current crop of non-peptide GLP-1 receptor agonists achieves it by binding a site that overlaps only partially with the peptide binding pocket, stabilising an active conformation without the two-domain capture mechanism.

Pharmacologically this matters for three reasons. Absorption does not depend on a permeation enhancer, so bioavailability is far less variable and far less dependent on fasting state than oral semaglutide’s. Elimination is hepatic rather than largely renal and proteolytic, which changes the interaction profile. And potency at the receptor is achieved without a fatty-acid albumin depot, so the concentration-time profile looks like a conventional small molecule rather than a peptide. None of this predicts efficacy; all of it predicts a different practical drug.

Why any of this belongs in a general publication

An argument could be made that receptor pharmacology is a specialist concern and that readers need practical guidance instead. The Journal’s position is the opposite, for a specific reason: almost every piece of bad advice circulating about this drug class is a mechanistic error with a practical conclusion attached.

Escalating on a fixed calendar regardless of symptoms is an error about accumulation kinetics. Splitting a weekly dose into daily fractions to reduce side effects is an error about half-life and steady state. Assuming a molecule with GIP activity is simply a stronger version of one without is an error about selectivity. Expecting weight to keep falling indefinitely is an error about energy balance. In each case the practical advice is wrong because the mechanism was misunderstood, and in each case understanding the mechanism is not much harder than memorising the rule.

A note on sources

Everything above is drawn from the peer-reviewed pharmacology and clinical literature and from regulatory assessment reports, which are more informative than the papers on questions of dose selection and exposure. Where a claim rests on in-vitro work in transfected cells, this piece says so, because the translation of such work to human physiology has failed often enough in this field to deserve a standing caveat.

Where the Journal reports a trial number it states the estimand behind it, because the treatment-policy and trial-product estimands differ by two to three percentage points in the obesity programmes and the difference is routinely lost in secondary coverage. Nothing here is a recommendation, and none of the compounds discussed as research chemicals are approved for human use.

A short glossary, because the words are used loosely

Agonist: a ligand that binds a receptor and produces a response. Full agonist: one producing the maximal response the system permits. Partial agonist: one producing less than maximal response even at full occupancy. Analogue: a molecule structurally derived from a natural ligand. Mimetic: a molecule reproducing a natural ligand’s effect without structural derivation.

Orthosteric site: the binding site the natural ligand occupies. Allosteric site: a distinct site whose occupancy modulates activity at the orthosteric one. Biased agonism: preferential activation of one downstream pathway over another. Tachyphylaxis: diminishing response to repeated administration. Steady state: the condition in which the rate of drug entering the body equals the rate leaving it.

Precision here is not pedantry. Several of the arguments this publication receives by post turn out, on inspection, to be disagreements about which of these words the writer meant.

What has actually changed in the last three years

Three things, on the Journal’s assessment. First, the demonstration that a dual agonist could produce weight reduction approaching bariatric-surgical magnitude moved the field’s expectations, and with them the design of every subsequent programme. Second, the cardiovascular and renal outcome results reframed the class from metabolic-cosmetic to cardiometabolic, which changed reimbursement arguments far more than it changed prescribing.

Third, and least remarked, the pharmacology of oral administration became tractable. That is a manufacturing and access story as much as a scientific one: an oral small molecule has a completely different cost structure, cold-chain requirement and supply profile from an injectable peptide, and if it holds up in phase 3 it will do more to change who can get treated than any of the receptor science described above.

Readers sometimes ask why a publication covering a consumer-facing drug class spends this much space on binding kinetics. The answer is that the alternative is a publication that reprints press releases with adjectives added. Mechanism is the only defence against that, and it is available to anybody willing to read a figure legend.