Semaglutide hits one receptor. Tirzepatide hits two. Retatrutide hits three. The pattern looks like a simple ladder, and the trial results roughly follow it — but the reason why is more interesting, and more instructive, than the headline numbers.
The incretin field has moved faster than almost any other area of metabolic pharmacology in the last decade, and the compounds now under study make a useful case study in a broader principle: what happens when you deliberately design a molecule to be less selective.
The receptors involved
Three class B G-protein-coupled receptors carry most of the story.
GLP-1R binds glucagon-like peptide-1, released from intestinal L-cells after a meal. Activation potentiates glucose-dependent insulin secretion, suppresses glucagon, slows gastric emptying and acts on hypothalamic circuits involved in food intake. The glucose dependence matters: GLP-1R agonism amplifies insulin release when glucose is high and largely stops when it is not.
GIPR binds glucose-dependent insulinotropic polypeptide, released from intestinal K-cells. Its role is the most contested of the three. GIP is insulinotropic, but it also acts on adipose tissue, and its signalling in the central nervous system appears to influence food intake through pathways partly distinct from GLP-1. The contest is over direction, and we will come back to it.
GCGR binds glucagon — the hormone usually cast as insulin's opposite. Glucagon raises hepatic glucose output, which sounds like exactly what you would not want in a metabolic agent. It also increases energy expenditure and promotes hepatic lipolysis. The bet in adding glucagon agonism is that the energy expenditure and lipid effects can be captured while the glycaemic downside is offset by simultaneous GLP-1 activity.
Why unimolecular multi-agonism is hard
Combining two drugs is straightforward. Building one peptide that engages several receptors at deliberately chosen relative potencies is not.
The starting point is helpful: GLP-1, GIP and glucagon are all members of the same peptide family and share substantial sequence homology, which is why a single backbone can be tuned toward several of them. The difficulty is that every substitution that shifts affinity at one receptor tends to shift it at the others, and the intended ratio has to survive all of it.
Half-life is a second constraint. Native GLP-1 is cleared in minutes, largely by DPP-4. Weekly dosing requires resistance to that cleavage — typically a substitution at position 2 — plus a strategy for albumin binding, usually a fatty-acid chain attached through a linker. Semaglutide's C18 diacid and linker are the well-known example. That modification also has to be compatible with the receptor pharmacology you spent the rest of the molecule building.
The result is that these compounds are not "GLP-1 plus GIP" in any additive sense. They are single molecules with a specific potency ratio at each receptor, and that ratio is a design decision that distinguishes one candidate from another as much as the receptor list does.
The GIP puzzle
The most genuinely interesting problem in the field is that both GIPR agonism and GIPR antagonism appear to produce metabolic benefit, which should not be possible if the mechanism were simple.
Tirzepatide is a GIPR/GLP-1R dual agonist and outperformed selective GLP-1R agonism in head-to-head trials. Meanwhile, antibody-based GIPR antagonists combined with GLP-1R agonism have also shown benefit in preclinical and early clinical work. Both cannot be straightforwardly true of the same pathway.
Several explanations are under active investigation. One is receptor desensitisation: sustained agonism at GIPR may functionally downregulate signalling, so a long-acting agonist and an antagonist converge on a similar end state by different routes. Another is tissue divergence — central and peripheral GIPR populations may contribute in opposite directions, so the net effect depends on where the compound acts and how much reaches the brain. A third is biased agonism, where different ligands at the same receptor recruit G-protein and β-arrestin pathways in different proportions and produce genuinely different downstream outcomes.
This is worth dwelling on because it is a good corrective to the ladder framing. "More receptors is better" is a summary of some results, not a mechanism. The field does not currently have a settled account of why GIPR engagement helps.
Adding glucagon
Retatrutide extends the approach to GCGR. The rationale is that glucagon's effect on energy expenditure and hepatic lipid handling adds something the incretins do not provide, particularly for hepatic steatosis, while concurrent GLP-1R agonism restrains the rise in hepatic glucose output.
The pharmacological tension is real and the balance is delicate. Too much glucagon agonism relative to GLP-1 and glycaemic control suffers; too little and there is no point adding it. This is precisely where the potency ratio, rather than the receptor list, does the work — and it is why triple agonists from different programmes are not interchangeable even when described identically.
What research models can and cannot show
A few methodological points recur in the literature and are worth stating plainly.
Species differences are substantial. Rodent and human GIPR and GCGR differ enough in sequence and expression that potency ratios established in one do not transfer cleanly to the other. Compounds are routinely characterised against both human and rodent receptors for this reason.
In-vitro potency is not in-vivo effect. A cAMP accumulation assay in a cell line overexpressing a single receptor measures one thing well. It does not capture receptor reserve, tissue distribution, competing endogenous ligands or pharmacokinetics. EC50 ratios from such assays are a starting point for a hypothesis, not a prediction.
Body-composition endpoints need composition data. Weight change alone cannot distinguish fat loss from lean mass loss, and the distinction is central to evaluating this class. Studies that report only mass are answering a narrower question than they appear to.
Where the field appears to be heading
Two directions look most active. The first is amylin: cagrilintide is a long-acting amylin analogue acting through a distinct satiety pathway, and combination with semaglutide is under investigation as CagriSema. The second is oral delivery, where the obstacle is straightforward — peptides are digested — and the solutions involve absorption enhancers and small-molecule agonists that are not peptides at all.
The broader lesson generalises past this class. For decades, selectivity was the goal in drug design and off-target activity was a defect. Incretin multi-agonists invert that: the polypharmacology is the mechanism, and the engineering problem is controlling it precisely rather than eliminating it. Whether that approach travels to other receptor families is one of the more interesting open questions in peptide pharmacology.
Reference material for the compounds discussed here is on their individual product pages: semaglutide, tirzepatide, retatrutide and cagrilintide.
All products are supplied strictly for in-vitro laboratory research use. Not for human or veterinary use. Nothing here is a recommendation to use any compound in a person or an animal.