Semaglutide — N-terminal conjugation of a short discrete PEG2 (8-amino-3,6-dioxaoctanoic acid, AEEA) spacer to the α-amine of His-1 via a stable amide bond, yielding AEEA-HAEGTFTSDVSSYLEGQAAK(γGlu-γGlu-C18diacid)EFIAWLVRGRG. Native Aib-2 and Lys-20 lipidation are preserved.

Semaglutide is a benchmark GLP-1 receptor agonist engineered for once-weekly subcutaneous dosing through two structural innovations: an Aib substitution at position 2 (blocking DPP-IV cleavage) and a γGlu-γGlu-C18 fatty diacid chain at Lys-26 (conferring high albumin affinity and a ~46-hour half-life). Despite these modifications, oral semaglutide (Rybelsus) achieves only ~1% bioavailability even in the presence of the absorption enhancer SNAC, a limitation driven in part by luminal aminopeptidase degradation of the exposed His-1 N-terminus. Fold №81 addresses this bottleneck directly by proposing the conjugation of a discrete PEG2 (AEEA, 8-amino-3,6-dioxaoctanoic acid) spacer to the α-amine of His-1 via a stable amide bond — a delivery-focused N-terminal modification not previously explored in this lab's semaglutide series.

The modification rationale is specific and layered. AEEA is a short (~10 Å), rigid-flexible discrete PEG unit that is substantially less immunogenic and sterically disruptive than polymeric PEG chains. By masking the α-amine of His-1 with an amide bond to AEEA, the hypothesis predicts that the principal aminopeptidase recognition site is occluded, reducing the rate of N-terminal proteolytic cleavage in gastric and mucosal environments. The AEEA unit is short enough to project away from the orthosteric binding pocket into bulk solvent, leaving the His-1 imidazole side chain — the critical pharmacophore for TM6 engagement in GLP-1R — geometrically accessible. The Aib-2 and Lys-26 lipidation elements are preserved unchanged, meaning albumin binding and DPP-IV resistance should be unaffected by the N-terminal addition.

However, there is a fundamental chemical tension at the core of this hypothesis. Cryo-EM structures of GLP-1 bound to GLP-1R consistently show that the α-amine of His-1 forms a direct hydrogen bond to Glu387 in TM6. The AEEA amide conjugation replaces the primary α-amine (pKa ~8, positively charged at physiological pH, strong H-bond donor) with a secondary amide nitrogen (non-protonatable, weakened H-bond donor capacity). This is not a trivial perturbation — it removes a charged pharmacophoric element that makes a specific polar contact in the transmembrane bundle. The literature supports the view that the mid-helical and C-terminal regions of GLP-1 engage the ECD, while the N-terminal residues 1-6 are indispensable for receptor activation: this has been called the 'two-domain' model of GLP-1R activation. Modifications distal to His-1 (as in the Lys-26 lipidation) are tolerated; modifications at His-1 itself are high-risk for potency.

The structural prediction from Boltz-2 returned a pLDDT of 0.68, a pTM of 0.66, and critically an ipTM of 0.46. The ipTM score is the primary metric for complex confidence in AlphaFold-family models, and values below ~0.50 are generally considered unreliable for structural interpretation. No Chai-1 ensemble agreement was obtained, and the Boltz-2 affinity module did not return binding change values. These metrics together constitute a tool-limit failure rather than a biological verdict: the model could not confidently place the AEEA-extended N-terminus relative to the GLP-1R binding interface, likely because (a) the AEEA linker introduces non-canonical chemistry that is poorly represented in training data, and (b) the N-terminal binding geometry in Class B GPCRs involves a narrow transmembrane cavity that is difficult to model with high confidence even for native peptides. The heuristic sequence-based profile (aggregation propensity 0.19, stability score 0.20, BBB penetration 0.11) is consistent with a hydrophilic, albumin-bound peptide with low CNS penetration — unremarkable for this class.

This fold connects meaningfully to prior semaglutide work in the lab. Fold №52 tested α-methyl-L-histidine at position 1 — a Cα-methylation that rigidifies the N-cap conformation while preserving the imidazole — and returned a PROMISING verdict (pLDDT 0.72, ipTM above threshold). The contrast is instructive: Cα-methylation at His-1 modifies the backbone geometry without altering the α-amine charge state, whereas AEEA conjugation converts the α-amine into an amide, a chemically more disruptive change. Fold №75 tested C-terminal truncation (PROMISING, pLDDT 0.78), and Fold №15 tested a mid-helix homoglutamate substitution (PROMISING, pLDDT 0.71) — both modifications distal to the receptor-critical N-terminus and both evaluable by current tools. Fold №36 was discarded (pLDDT 0.70, poor ipTM) for the β-Ala linker substitution at the Lys-26 lipidation site, another modification involving non-canonical linker chemistry that challenged structural prediction tools.

The clinical agent context is important: semaglutide is already approved for both subcutaneous and oral routes, and the PIONEER trials have demonstrated meaningful glycemic and weight outcomes with oral dosing despite the low bioavailability. The incremental benefit of N-terminal AEEA modification must therefore be weighed against the real risk of potency reduction at GLP-1R — a concern amplified by the fact that oral semaglutide is dosed at 14 mg (vs. 0.5-2 mg for subcutaneous) precisely because bioavailability is low. If AEEA conjugation reduces receptor affinity by even 5-10 fold (plausible given α-amine conversion to amide), the formulation advantage could be offset.

This fold is discarded as a tool-limit result. The hypothesis remains scientifically novel — no published study has assessed N-terminal AEEA conjugation on semaglutide or any close GLP-1 analogue — and the delivery rationale is well-founded. What is needed is not better computational prediction but wet-lab functional assay: a cAMP accumulation assay or GLP-1R binding competition assay on the AEEA-His-1 conjugate would definitively answer whether the α-amine-to-amide conversion ablates, reduces, or preserves receptor activation. The AEEA peptide can be synthesized by standard Fmoc SPPS with an AEEA building block, making this experimentally accessible.