Multi-Receptor Peptide Design: One Molecule, Several Targets

July 3, 2026 14 min read The Human Machine

peptidespharmacologydrug-designgpcrbiotech

Contents

A receptor doesn’t care what you call the molecule that binds it — only what that binding does to the receptor’s shape. That distinction is the entire basis of modern peptide drug design: agonist, antagonist, and the newer, stranger category of biased agonist are not different kinds of molecules, they’re different descriptions of what happens to one receptor’s conformation and downstream signaling when something docks into it. Once you take that framing seriously, the next design move follows almost naturally — if a single peptide can be tuned to do one specific thing to one receptor, there is no fundamental reason it can’t be tuned to do specific, independently chosen things to two or three receptors at once. Tirzepatide (GLP-1 plus GIP), retatrutide (GLP-1 plus GIP plus glucagon), and MariTide (GLP-1 agonism paired with GIP antagonism in the same molecule) are not incremental tweaks on semaglutide. They’re a different design philosophy, and understanding why it works requires starting with what agonism and antagonism actually mean at the molecular level.

GLP-1 agonists, honestly covered semaglutide and tirzepatide’s clinical effect sizes and side-effect profiles in detail. This post is about the design logic underneath those drugs — the receptor pharmacology that explains why hitting more than one target with a single molecule became the industry’s dominant strategy, rather than just a headline-grabbing variation on a working idea.

Agonist, Antagonist, and the Space Between

The textbook framing is simpler than the reality. An agonist binds a receptor and triggers its normal downstream signal — GLP-1 binding the GLP-1 receptor activates a G protein cascade that raises intracellular cAMP, ultimately triggering insulin secretion. An antagonist binds the same receptor pocket, or an overlapping one, but produces no signal itself, and by occupying the site it physically blocks the natural ligand from binding — the receptor goes quiet, or at least quieter than baseline.

But most receptors have a resting, low-level “leak” of signaling even with nothing bound, and that resting signal turns “antagonist” into a category with two subtypes. A neutral antagonist blocks the natural ligand without changing the receptor’s baseline signal. An inverse agonist goes further, actively suppressing the receptor below its resting signaling level — useful when the therapeutic goal isn’t just “stop the natural ligand” but “push the pathway below its unstimulated baseline.” A partial agonist sits in between agonist and antagonist entirely: it activates the receptor, but with a lower maximum signal than the full natural ligand achieves, even at saturating concentration — which makes it useful as a stabilizer that prevents wild swings in either direction, blocking the full agonist from over-activating the receptor while still providing enough baseline signal to avoid full shutdown.

RECEPTOR SIGNALING OUTPUT
(relative to unstimulated baseline)

  Full agonist    ────────────────────────●  (max signal)
                                          /
  Partial agonist ─────────────●         /
                               /         /
  [baseline, no ligand] ──────●─────────────────────
                          \
  Neutral antagonist ──────●  (holds at baseline, blocks natural ligand)
                          \
  Inverse agonist ──────────●  (below baseline)

  low <---------------- ligand efficacy ----------------> high

None of this is exotic — it’s the framework every receptor pharmacology course teaches for small-molecule drugs too. What makes it relevant to peptides specifically is that peptide ligands are unusually easy to re-engineer into any point on that spectrum, because a receptor’s binding pocket typically recognizes a short, specific stretch of the peptide’s sequence. Change a handful of residues and you can shift a molecule from full agonist toward partial agonist, or from agonist to antagonist outright, while barely touching its overall size or manufacturability. That tunability is the raw material multi-receptor design builds on.

Biased Agonism: Same Receptor, Different Signal

The agonist/antagonist spectrum above assumes a receptor has one signaling output to turn up or down. Most G-protein-coupled receptors (GPCRs) — the family GLP-1, GIP, and glucagon receptors all belong to — actually have at least two parallel, mostly independent output pathways: classic G-protein signaling, and a separate pathway through proteins called beta-arrestins, which get recruited to the receptor after G-protein activation and typically handle receptor desensitization and internalization, but also carry their own independent signaling consequences.

Biased agonism is the discovery that a ligand’s shape can favor one of these pathways over the other, rather than activating both proportionally the way the receptor’s natural ligand does. The clearest approved example isn’t a peptide at all — it’s oliceridine (Olinvyk), a small-molecule opioid approved in 2020 that binds the mu-opioid receptor and strongly activates G-protein signaling (linked to analgesia) while recruiting beta-arrestin at only about 14% of the efficiency of morphine. Since beta-arrestin recruitment at the mu-opioid receptor is mechanistically tied to several of opioids’ worst adverse effects, a molecule that gets the pain relief without proportionally getting the arrestin-linked liability is a real, clinically meaningful improvement built entirely from a signaling-pathway distinction that a simple agonist/antagonist framework can’t capture.

The same logic applies to peptide GPCR targets, and it matters for the metabolic drugs this post is really about. Both G-protein and beta-arrestin signaling contribute to GLP-1 receptor agonists’ effects, but they don’t contribute equally to every downstream outcome — insulin secretion leans more heavily on the G-protein arm, while receptor desensitization (which limits how long a dose keeps working) is arrestin-driven. That’s part of the mechanistic argument for why semaglutide and tirzepatide, which differ somewhat in exactly how they engage each pathway, don’t behave identically dose-for-dose even accounting for their different receptor targets. Biased agonism doesn’t get its own drug class the way dual and triple agonism does yet, but it’s the pharmacological vocabulary underneath why receptor engineering choices that look small on paper — a few substituted residues — can change a drug’s clinical behavior more than the raw agonist/antagonist label would predict.

Why Hit Two Receptors Instead of One

The dual- and triple-agonist strategy didn’t emerge from an abstract desire for complexity. It emerged from a specific observation: hitting a second metabolically relevant receptor alongside GLP-1 produced effects that were more than additive, not just two half-strength drugs stapled together.

Tirzepatide agonizes both the GLP-1 receptor and the GIP receptor (glucose-dependent insulinotropic polypeptide receptor) in a single 39-amino-acid peptide. GIP, on its own, was historically considered the weaker sibling incretin — its insulin-secreting effect is blunted in people with type 2 diabetes in a way GLP-1’s isn’t, which for years made GIP receptor agonism look like a dead end rather than a target worth pursuing. Combining GIP agonism with GLP-1 agonism in the same molecule reversed that read entirely: GIP receptor activation appears to synergize with and amplify GLP-1’s effects on insulin secretion, and clinically, GIP co-agonism has also been associated with somewhat better tolerability at comparable efficacy than GLP-1 agonism alone, likely by modulating the nausea signal that GLP-1 agonism produces on its own.

Retatrutide pushes the same idea one receptor further, adding glucagon receptor agonism to the GLP-1/GIP combination. Glucagon receptor activation is, on its face, a strange choice to pair with drugs whose entire selling point is lowering blood glucose — glucagon’s classic physiological role is telling the liver to release stored glucose. But at the doses and in the combination retatrutide uses, glucagon receptor agonism drives increased energy expenditure and hepatic fat oxidation rather than net glucose elevation, and in phase 2 and phase 3 trials that third mechanism has translated into some of the largest weight-loss and liver-fat-reduction effect sizes reported for any peptide therapeutic to date:

Drug	Receptors targeted	Reported weight loss (pivotal trial data)	Notable secondary effect
Semaglutide	GLP-1 (agonist)	~15% average body weight	—
Tirzepatide	GLP-1 + GIP (both agonist)	~21% average body weight	Improved GI tolerability vs. GLP-1-only at comparable efficacy
Retatrutide	GLP-1 + GIP + glucagon (all agonist)	~24% average body weight at 48 weeks, 12mg dose	Up to ~82% reduction in liver fat in phase 2
MariTide	GLP-1 (agonist) + GIP (antagonist)	Up to ~20% average body weight at 52 weeks, phase 2	Monthly dosing feasible due to antibody-conjugate half-life

The pattern across the first three rows is straightforward: more receptors, engineered to work together rather than merely in parallel, produced larger effects. MariTide is the row that breaks the naive “more agonism is better” reading of that pattern, and it’s the more interesting design case for exactly that reason.

The Paradox Case: An Agonist and an Antagonist in the Same Molecule

MariTide (maridebart cafraglutide, developed by Amgen) pairs GLP-1 receptor agonism with GIP receptor antagonism — blocking the same receptor that tirzepatide activates — in a single bispecific molecule: an antibody fragment engineered to antagonize the GIP receptor, with two GLP-1 agonist peptides chemically conjugated onto it via amino acid linkers.

This looks like a direct contradiction of the tirzepatide logic until you look at the preclinical genetics that motivated it. Human genetic studies had found that loss-of-function variants in the GIP receptor gene were associated with lower body weight — the opposite direction you’d expect if GIP agonism were purely beneficial for weight loss on its own. That contradiction (GIP agonism helps in tirzepatide’s combination, but GIP loss-of-function also correlates with leanness) is exactly the kind of receptor-biology paradox multi-mechanism engineering exists to resolve empirically rather than theoretically: Amgen’s preclinical work found that combining GLP-1 agonism with GIP antagonism produced stronger weight loss in animal models than targeting either receptor alone in either direction, suggesting the relationship between GIP receptor activity and body weight isn’t monotonic — at least in the presence of concurrent GLP-1 agonism, blocking GIP appears to help rather than hurt, even though activating GIP alongside GLP-1 also helps. Both tirzepatide’s and MariTide’s phase 2 data are real, and the field’s honest current position is that both approaches work, for reasons the field does not yet have a fully unified mechanistic account of — a genuinely unresolved research question, not a settled one being simplified for a blog post.

The practical payoff of MariTide’s design has nothing to do with the agonist/antagonist question directly: the antibody-conjugate format gives it a long enough circulating half-life to support monthly dosing, versus the weekly injections tirzepatide and semaglutide require, which is a meaningful adherence advantage independent of whichever GIP-direction mechanism turns out to matter most.

Engineering One Peptide to Hit Several Targets

Building a peptide that reliably agonizes two or three distinct receptors isn’t a matter of stapling two separate drugs together — a single peptide backbone has to fold and present a distinct binding surface to each receptor without steric interference between the segments doing that work. Tirzepatide’s sequence is built on the native GIP backbone with strategic substitutions that add GLP-1 receptor affinity to the same fold, plus a fatty-diacid side chain (using the same albumin-binding lipidation trick covered in the peptide half-life problem) to stretch its half-life to roughly five days, supporting once-weekly dosing. Retatrutide extends the same backbone-engineering approach one receptor further, and MariTide sidesteps the single-backbone constraint entirely by using a fundamentally different format — an antibody scaffold with peptide agonists attached as separate conjugated arms rather than one continuous sequence trying to satisfy three receptor pockets at once.

A representative (illustrative, not clinical-grade) receptor-binding affinity screen for a multi-agonist candidate during early development looks like this — cross-reactivity has to be checked deliberately rather than assumed:

receptor_binding_panel:
  candidate: RA-0417
  assay: radioligand competition binding, HEK293 cells
          stably expressing each human receptor
  targets:
    - receptor: GLP1R
      Ki_nM: 0.8
      classification: full_agonist
    - receptor: GIPR
      Ki_nM: 2.1
      classification: full_agonist
    - receptor: GCGR   # glucagon receptor
      Ki_nM: 340.0
      classification: negligible_binding   # off-target check, want HIGH Ki here
    - receptor: GLP2R  # structurally related, must NOT cross-react
      Ki_nM: ">10000"
      classification: no_binding_detected
  functional_assay: cAMP accumulation, receptor-transfected cells
  notes: >
    Ki values below ~10 nM at intended targets confirm nanomolar-range
    affinity; Ki values above 1000 nM at structurally related off-targets
    (GLP2R, glucagon family) confirm adequate selectivity margin before
    proceeding to in vivo dosing.

That off-target selectivity check matters more for multi-receptor peptides than single-target ones, precisely because engineering a backbone to bind two or three intended receptors increases the chance it picks up incidental affinity for a fourth, unintended one from the same structurally related receptor family — GLP-1, GIP, glucagon, and glucagon-like peptide-2 receptors are all close evolutionary relatives with similar binding-pocket architecture, which is exactly why cross-reactivity screening against the near neighbors, not just confirmation of the intended targets, is a required step rather than an optional one.

The Trade-off Nobody Ships Around

Every receptor added to a peptide’s target profile is also a receptor whose side-effect profile now belongs to the drug. Glucagon receptor agonism’s main systemic risk is that, decoupled from concurrent GLP-1 suppression of hepatic glucose output, activating it could in principle raise blood glucose rather than lower it — which is precisely why retatrutide’s viability depends on the GLP-1 and GIP components doing enough concurrent work to keep that risk from manifesting clinically, not on the glucagon component being risk-free in isolation. This is the general shape of the trade-off multi-receptor design always makes: more receptors means more independent avenues for a therapeutic benefit, but also more independent avenues for something to go wrong, and the only way to know the net effect is actually favorable is the trial data itself, not the receptor logic on a whiteboard.

Honest Trade-offs

“More receptors” is not automatically “more effect.” GIP receptor agonism and GIP receptor antagonism have both shown weight-loss benefit in combination with GLP-1 agonism in different molecules (tirzepatide and MariTide respectively) — the field does not yet have a fully settled mechanistic explanation for why both directions work, which should temper confidence in extrapolating this design pattern to other receptor pairs without dedicated trial data for each combination.
Biased agonism is real pharmacology, not marketing language, but it doesn’t yet have its own approved peptide drug class. Oliceridine proves the concept works clinically for a small molecule at a GPCR; the metabolic peptide field’s biased-signaling differences are documented mechanistically but aren’t yet the basis of an approved drug marketed specifically on that property.
Each added receptor is an added manufacturing and regulatory burden, not just a pharmacology decision. A dual- or triple-agonist backbone, or a bispecific antibody-peptide conjugate like MariTide, is a harder synthesis and characterization problem than a single-receptor peptide, which is part of why these drugs have taken years longer to reach market than the single-target GLP-1 agonists that preceded them.
Larger effect sizes in trials have generally come with a broader side-effect surface to monitor, not a cleaner one. Retatrutide’s larger weight-loss numbers are real, but a third active receptor mechanism is also a third mechanism regulators and physicians have to independently rule out as a source of any adverse event that shows up.
The selectivity-screening burden scales with the number of intended targets. Confirming a peptide hits its two or three intended receptors is necessary but not sufficient — ruling out unintended affinity for structurally related receptors in the same family becomes proportionally more important, and more work, as the intended target list grows.

Verdict

Multi-receptor peptide design isn’t a single trick — it’s the natural extension of a fact that’s been true of GPCR pharmacology for decades (agonism, antagonism, and biased signaling are all just different descriptions of receptor conformational outcomes) applied to a class of molecules, peptides, that happen to be unusually easy to re-engineer at the sequence level. Tirzepatide and retatrutide show that combining agonism at several related receptors in one backbone can produce effects meaningfully larger than any single-receptor drug in the same family. MariTide shows the same design space is more complicated than “activate everything relevant” — sometimes the winning combination pairs agonism at one receptor with antagonism at a close relative, for reasons the field is still working out. What’s consistent across all three is that the receptor-level logic, not the headline weight-loss percentage, is what will keep generating the next generation of these drugs — and the next combination that works is as likely to come from a genetic paradox like GIP’s as from simply adding another agonist to the pile.