Semaglutide: Structure, Albumin Binding and Receptor Pharmacology

Semaglutide is human glucagon-like peptide-1 with two residues swapped and a fatty acid hung off a lysine. That’s it. Those three changes make it stick to albumin and shrug off the enzyme that normally destroys GLP-1 within minutes. We went back to the Novo Nordisk characterization paper and the structural work that followed it, because almost everything written about this peptide online is a summary of a summary. Tirzepatide comes into it too, since the two get compared constantly and the pharmacology really is different.

What the molecule is

Native GLP-1 barely lasts. It’s a 30 residue incretin and dipeptidyl peptidase-4 chews on its N-terminus almost immediately, so any long acting version has to beat both the enzyme and the kidney. Lau and colleagues at Novo Nordisk described how semaglutide was arrived at [1]. The Aib substitution at position 8 blocks the enzymatic cut. Arg34 removes an alternative acylation site so the fatty acid goes where it is meant to. The derivatization at lysine 26, a C18 diacid tied on through a short linker, is what gives the peptide its albumin affinity.

That design has a cost the paper states plainly. Receptor affinity dropped roughly threefold relative to liraglutide, to 0.38 nM [1]. Less potency, far more time in circulation. That bargain is routine in peptide engineering, and it’s why comparing raw affinity numbers between analogues as if they were quality scores tells you very little.

At the receptor

GLP-1R is a class B G protein coupled receptor. Zhang’s group solved cryo-EM structures of it bound to semaglutide and to taspoglutide, with Gs attached, then ran 3D variability analysis [2]. Contacts looked much like native GLP-1. The receptor itself did not behave the same way, and its internal motions shifted depending on which agonist was sitting in the pocket. Pharmacologists had been inferring exactly that from signalling assays for years. Here it is in a structure.

Distribution follows from the albumin binding. Gabery and colleagues mapped where semaglutide actually went in rodents and found it did not cross the blood brain barrier [5]. It got to the brainstem, septal nucleus and hypothalamus anyway, by way of the circumventricular organs and a few spots next to the ventricles. c-Fos lit up in ten areas. Some of those had no direct receptor interaction at all, so second order circuits are doing part of the work.

Semaglutide and tirzepatide side by side

People compare these two constantly and usually badly. The difference isn’t strength. It’s what each one touches.

Willard’s team worked out receptor occupancy for each target and described tirzepatide as imbalanced and biased, which is a precise claim, not a marketing one [3]. In primary islets, beta-arrestin1 capped the insulin response to GLP-1, yet left GIP and tirzepatide alone. Read out cAMP only and you miss the bit that separates them.

Stability, oligomerization and degradation

Lipidated peptides are strange in solution. Gallo’s group put NMR and surface plasmon resonance on liraglutide and semaglutide analogues, and found oligomerization and albumin binding pulling in different directions [4]. Oligomers were shielded from catabolism. Albumin helped less than you’d guess, since the unbound fraction runs higher in tissue than in plasma. Catabolite work in rat plasma put the main cut at Tyr19-Leu20, and the in vitro picture matched reasonably well.

Why care? Concentration sets aggregation state, aggregation state sets degradation rate. Two labs at the same nominal concentration in different buffers can get different stability numbers with neither of them doing anything wrong. Lyophilized material should be characterized on arrival rather than assumed, and the certificate of analysis is where to start; our notes on third party tested material cover what a useful report contains.

Absorption across the stomach wall

One of the more interesting pharmacokinetic papers on semaglutide has nothing to do with receptors. Buckley and colleagues studied the peptide coformulated with sodium N-[8-(2-hydroxybenzoyl) aminocaprylate], or SNAC, in tablet form [7]. In dogs and in clinical work, absorption happened in the stomach rather than the intestine, was confined to a small area close to the tablet surface, and did not occur without SNAC. The mechanism was transcellular with no evidence of tight junction disruption, and SNAC appeared to work partly by buffering locally against enzymatic degradation. It is a nice example of formulation chemistry doing something a peptide sequence cannot do on its own. Our overview of oral peptide delivery puts this in wider context.

Characterizing the material

A lipidated peptide is harder to analyse than a plain one, and the reason is the fatty acid. On reversed phase HPLC the C18 diacid dominates retention, so impurities that differ from the main peak only in the peptide portion can run very close to it. Resolving them takes a well developed gradient, and a chromatogram with one clean peak at a shallow gradient does not prove much. Orthogonal separation, pairing hydrophilic interaction chromatography with reversed phase, is the usual way around that problem for complex synthetic peptides.

Mass spectrometry has blind spots too. Isobaric substitutions and aspartic acid isomerization leave the mass untouched, so intact mass misses them entirely and you need peptide mapping after digestion. Reference standard practice leans on chromatographic content assignment plus mass spectrometry and NMR for identity, and keeps content uniformity and stability as questions separate from purity [6]. Then there’s the counterion, which everyone forgets. Trifluoroacetate salt and acetate salt at the same stated purity do not hold the same mass of peptide, and that error quietly propagates into every concentration you calculate afterwards.

A note on comparing the two peptides

One last caution, because this is where most write-ups go wrong. Semaglutide and tirzepatide were characterized by different companies, in different assay systems, with different reference ligands. Willard’s group calculated receptor occupancy from their own data; Lau’s affinity figure sits next to liraglutide, not next to tirzepatide [1][3]. Lifting a number from each paper and putting them side by side in a table is tempting and mostly meaningless, which is why the comparison table above lists receptor behaviour rather than head to head potency. A clean comparison would need both peptides run in one laboratory, in the same cells, on the same day. As far as we can tell from the published record, that study has not been done for the endpoints people most want compared.

Frequently asked questions

Is semaglutide a peptide or a small molecule?

A peptide. It is a modified 31 residue GLP-1 analogue with a fatty diacid chain attached at lysine 26 [1]. Orally available small molecule GLP-1 receptor agonists exist and are a different chemical class entirely.

Why does tirzepatide get called biased?

Because at the GLP-1 receptor it favours cAMP generation over beta-arrestin recruitment, rather than activating both arms in the proportions native GLP-1 does [3]. Bias is a specific pharmacological term describing the ratio of signalling outputs.

How long does semaglutide last in circulation?

In mini-pigs the plasma half-life was 46.1 hours after intravenous delivery [1]. Half-life is species dependent, so a rodent figure and a mini-pig figure are not interchangeable.

References

1. Lau J, Bloch P, Schaffer L, et al. Discovery of the Once-Weekly Glucagon-Like Peptide-1 (GLP-1) Analogue Semaglutide. J Med Chem. 2015. PMID 26308095
2. Zhang X, Belousoff MJ, Liang YL, Danev R, Sexton PM, Wootten D. Structure and dynamics of semaglutide- and taspoglutide-bound GLP-1R-Gs complexes. Cell Rep. 2021. PMID 34260945
3. Willard FS, Douros JD, Gabe MB, et al. Tirzepatide is an imbalanced and biased dual GIP and GLP-1 receptor agonist. JCI Insight. 2020. PMID 32730231
4. Gallo M, Vanni D, Esposito S, et al. Oligomerization, albumin binding and catabolism of therapeutic peptides: an investigation on lipidated GLP-1 analogs. J Pharm Biomed Anal. 2022. PMID 35042144
5. Gabery S, Salinas CG, Paulsen SJ, et al. Semaglutide lowers body weight in rodents via distributed neural pathways. JCI Insight. 2020. PMID 32213703
6. McCarthy D, Han Y, Carrick K, et al. Reference Standards to Support Quality of Synthetic Peptide Therapeutics. Pharm Res. 2023. PMID 36949371
7. Buckley ST, Baekdal TA, Vegge A, et al. Transcellular stomach absorption of a derivatized glucagon-like peptide-1 receptor agonist. Sci Transl Med. 2018. PMID 30429357