Design, Structure–Activity Relationships, and Computational Modeling Studies of a Series of α-Helix Biased, Ultra-Short Glucagon-like Peptide-1 Receptor Agonists
The GLP-1 drugs in clinical use are large peptides carrying fatty acid chains. This work takes an eleven-residue agonist apart position by position, and finds that much of its potency comes not from what its side chains touch, but from forcing the backbone into a helix before it arrives.
Order bought in advance: the study's central finding is that constraining the backbone into a helix can substitute for some of what the side chains do.
The GLP-1 drugs that have reshaped the treatment of type 2 diabetes and obesity are, chemically speaking, large. Liraglutide and semaglutide are built on the GLP-1 (7–37) fragment and hung with fatty acid chains that bind albumin, creating a reservoir in plasma and stretching a native half-life of a few minutes into weekly dosing. It is an elegant solution, and it has worked.
It also raises an obvious question. Native GLP-1 is thirty-odd residues, but only the N-terminal stretch inserts into the receptor’s active site; the rest engages the extracellular domain. How much of the molecule is actually doing the pharmacology — and how short could a GLP-1 receptor agonist be?
Researchers at Bristol Myers Squibb answered the first version of that question some years ago, showing that ultra-short eleven-residue peptides could reach potencies comparable to native GLP-1. A team spanning Resolute Bio, Eudoxia Life Sciences, XtalPi, the University of Arizona, Peptide Scientia and Maestro Therapeutics returned to that scaffold to ask a different one: why do these short peptides work, and what exactly are the modifications doing?
Their answer, published in Molecules, is largely about shape — specifically about persuading a very short chain to hold a helix before it ever reaches the receptor.
Eleven residues
The template is H-His-Aib-Glu-Gly-Thr-Phe-Thr-Ser-Asp-Bip-Bip-NH2: the first nine positions recognisably GLP-1, then two biphenylalanine residues standing in for the rest of the hormone. Two unusual building blocks appear throughout. Aib — 2-aminoisobutyric acid — is the Cα-methylated cousin of alanine used in liraglutide and semaglutide at position 2, where it blocks cleavage by DPP-4. Bip is p-phenyl-phenylalanine, a large hydrophobic residue.
The team worked through the series systematically: fourteen analogues at position 6 alone, then a full alanine scan and a full Aib scan across the eleven positions, with potency measured as cAMP EC50 at the GLP-1 receptor in an HTRF assay.
At position 6, the simplest change proved the best. Substituting Phe with Phe(2-F) — a single fluorine on the ring — gave roughly a fivefold gain over the parent. Other substituents at the same position were markedly worse: Phe(2-Cl), Phe(2-CF3), Phe(2-Me) and Phe(NO2) all cost between ten and a hundred fold. Phe(2,6-F) was about equipotent; the fully pentafluorinated version was threefold weaker. Small changes, large consequences, and not in a pattern that simple bulk or electronegativity would predict.
Reading two scans against each other
The more interesting result comes from running the alanine and Aib scans side by side. Alanine substitution asks what a side chain contributes. Aib substitution asks the same question while also adding a Cα-methyl group, which restricts the backbone’s freedom and biases it toward a helix. The difference between the two scans at a given position therefore isolates what the conformational constraint is worth.
The alanine scan identified the load-bearing residues: His1, Gly4 and Phe6 in the N-terminal region, and both Bip residues at positions 10 and 11, each costing more than a thousandfold when replaced. The Aib scan told a different story in places. At position 4, Aib recovered about tenfold of an otherwise two-thousandfold loss. At position 6, Aib rescued more than tenfold of a greater-than-thousandfold loss.
Position 11 was the most striking. Losing Bip11 to alanine cost around ten-thousandfold; replacing it with Aib recovered more than a hundredfold of that. The authors read this as the constraint doing real work — that the enhanced helicity induced by Aib11 partly compensates for the enthalpic interactions lost when the Bip side chain goes. Shape, in other words, can substitute for contact.
Quantifying this, they estimate the contribution of Cα-methylation at position 6 at 0.97 pEC50 units, derived from the difference between the two parent peptides in the series, and use that figure as the baseline against which methylation elsewhere is judged.
The enhanced helicity induced by Aib11 alleviated the potency impact caused by a loss of enthalpic interactions of Bip11 at the GLP-1R.
A model, and its limits
The structural half of the work builds on published GLP-1 receptor structures — X-ray and cryo-EM, including complexes with full-length GLP-1 and with a truncated peptide agonist — to construct a binding model for this series, and from it a quantitative structure–activity relationship.
The final QSAR relates potency to computed binding energy terms and accounts for roughly 60% of the variation in the measured pEC50 data. The authors are candid about the constraints: earlier regressions in the paper carry weak cross-validated statistics, and they note that a model should not be expected to outperform the reproducibility of the assay it is trained against. Two weighted terms exceed the conventional rule of thumb for the number of data points available. It is presented as a working hypothesis-generating tool rather than a predictive instrument.
What it adds
The authors frame the work as recapitulating and extending the BMS findings rather than superseding them, and that is the right reading. What is new is the systematic double scan and what it reveals: a per-position accounting of how much of this scaffold’s potency comes from side-chain contact and how much from backbone conformation.
Two practical conclusions fall out. His1 can be replaced by either Ala1 or Aib1 — a position long assumed critical in the full-length hormone proves more forgiving here. And a template carrying Bip10 with Aib11 yields a low-nanomolar agonist, suggesting the C-terminal anchors can be traded partly for constraint.
This is in vitro pharmacology on a receptor assay. The paper reports potency and efficacy at the GLP-1 receptor; it does not address proteolytic stability, pharmacokinetics, half-life or in vivo activity, and makes no therapeutic claim. Whether an eleven-residue agonist can be given to a patient is a question about properties this study did not measure.
What it does offer is a design principle with numbers attached. If a short peptide’s problem is that it pays an entropic price to fold on binding, then methylating the right backbone positions makes it fold in advance — and the scans here say where, and roughly how much it is worth.
Source
Jonathon R. Sawyer, Joseph A. Audie, Jon Swanson, David Diller, Solimar Santiago, Valentin K. Gribkoff, Allison Ackerman, Victor J. Hruby, Gianpaolo Gobbo, Michael A. Bellucci, William A. Glauser, Brad L. Pentelute and Tomi K. Sawyer. “Design, Structure–Activity Relationships, and Computational Modeling Studies of a Series of α-Helix Biased, Ultra-Short Glucagon-like Peptide-1 Receptor Agonists.” Molecules, volume 30, article 12. Received 5 April 2024; revised 24 July 2024; accepted 29 July 2024; published 24 December 2024. MDPI, Basel. DOI: 10.3390/molecules30010012. Open access under CC BY 4.0.
Affiliations: Resolute Bio, Beverly, Massachusetts; Department of Chemistry and Biochemistry, The University of Arizona; Peptide Scientia, Southborough, Massachusetts; Eudoxia Life Sciences, Cheshire, Connecticut; XtalPi US, Cambridge, Massachusetts; Maestro Therapeutics, Southborough, Massachusetts. Jonathon R. Sawyer and Joseph A. Audie contributed equally and are the corresponding authors.
The underlying research is the work of the cited authors and their institutions. Maestro Chronicle provides the editorial presentation.