Fold №50 — Tesamorelin Hexa-Position Hydrocarbon Staple (Val-13/Leu-17 → S5 i,i+4)

Verdict: DISCARDED | pLDDT 0.46 | ipTM 0.17

Mechanism of Action (Background)

Tesamorelin is a 44-amino acid synthetic analogue of human growth hormone-releasing hormone (GHRH), carrying an N-terminal trans-3-hexenoyl pharmacophore that protects against DPP-IV cleavage at Ala-2. It acts as a full agonist at the GHRHR (UniProt Q02643), stimulating pulsatile GH secretion from anterior pituitary somatotrophs. The GH pulse drives downstream IGF-1 production, lipolysis in visceral adipose tissue, and the hepatic/metabolic endpoints that make Tesamorelin clinically valuable in HIV-associated lipodystrophy.

The peptide's central segment (residues ~9–22 of the GRF(1-29) core) is predicted to adopt an amphipathic α-helix that docks into the GHRHR extracellular domain. In solution, however, the free peptide samples substantially disordered states, potentially reducing binding affinity relative to the bound-state conformation. The trans-3-hexenoyl cap provides N-terminal stability but does nothing to pre-organize the central helix — this is the conformational vulnerability the current fold aimed to address.

Modification Hypothesis (What We Tested)

This fold installed a single i,i+4 all-hydrocarbon staple between positions 13 (Val → (S)-2-(4'-pentenyl)alanine, S5) and 17 (Leu → S5), cross-linked via ruthenium-catalyzed ring-closing metathesis to form a covalent macrocyclic bridge across the putative solvent-exposed face of the central helix. The hypothesis was that this covalent lock would:

1. Pre-organize the bioactive α-helical conformation, reducing conformational entropy upon GHRHR binding and potentially increasing binding affinity
2. Confer protease resistance along the central helix, complementing the N-terminal trans-3-hexenoyl cap
3. Preserve the N-terminal pharmacophore and C-terminal recognition motif, since the staple was placed on a solvent-exposed hydrophobic face away from predicted receptor contact residues

The approach was deliberately designed to avoid the Aib substitution chemistry that failed in Fold №13 (Gln-8 → Aib, pLDDT 0.49, DISCARDED) and Fold №29 (Ala-2 → Aib, pLDDT 0.47, DISCARDED), and to address a different failure mode — those modifications nucleate helicity locally but cannot covalently enforce it across the full helix span.

Why the Prediction Was Uninformative (Technical Analysis)

The core technical problem is that (S)-2-(4'-pentenyl)alanine (S5) residues are entirely outside AlphaFold2's training distribution. The model has no learned representations for unnatural amino acids with extended aliphatic side-chains and covalent ring-closing cross-links. Presented with two S5 positions and an implicit covalent bridge, the predictor almost certainly interprets these residues as disordered noise, defaulting to a low-confidence unstructured prediction rather than extrapolating the helical geometry that the physical staple would enforce.

This is the same barrier that produced the DISCARDED verdict in Fold №42 (Sermorelin Lys-21/Asp-25 i,i+4 lactam staple, pLDDT 0.50, DISCARDED). That fold involved a different staple chemistry (lactam vs. hydrocarbon) and a different peptide, but the structural predictor responded identically — with a low-confidence, interface-uninformative output. Together, Folds №42 and №50 establish a reproducible empirical finding: current AF2-based tools cannot evaluate conformational-lock strategies on GHRHR-targeting peptides involving non-canonical cross-linking chemistry.

The heuristic sequence-based profile (aggregation propensity 0.141, stability score 0.378, long half-life estimate) provides no useful signal here because it does not account for the actual staple geometry, cross-link rigidity, or the protease-shielding effect of the hydrocarbon bridge. These numbers describe the underlying Tesamorelin sequence, not the stapled analogue.

What This Tells Us (Negative Results Are Data)

This fold rules out one thing confidently: in silico triage of hydrocarbon-stapled Tesamorelin analogues is not currently feasible with this prediction stack. This is a meaningful finding for lab workflow — it means any future stapled GHRH-family peptide must be routed directly to wet-lab validation rather than filtered computationally.

It does not tell us that the staple concept is chemically flawed. The scientific rationale — covalent helix pre-organization at positions on the solvent-exposed face — remains mechanistically plausible and draws on a well-validated general framework (Verdine/Walensky chemistry, clinical precedent in ALRN-6924). The absence of an interpretable in silico signal is not evidence of failure in the molecule; it is evidence of a tool capability gap.

The broader pattern across Tesamorelin folds (№13, №29, №50, all DISCARDED) and the related Sermorelin staple (Fold №42, DISCARDED) establishes a lab-level insight: GHRHR-targeting long-chain peptides with non-canonical backbone modifications consistently defeat current structure predictors. Folds targeting smaller peptides with more conventional modifications — such as Ipamorelin with palmitoyl lipidation (Fold №48, REFINED, pLDDT 0.78) or Ipamorelin macrocyclization (Fold №33, REFINED, pLDDT 0.73) — do produce meaningful signal. The predictor succeeds on pentapeptide GHSR-1a ligands with compatible chemistry; it fails on 44-mer GHRHR ligands with non-canonical cross-links.

Alternative Hypotheses to Test (Avoiding the Failure Mode)

Route A — Wet-lab bypass (recommended for staple concept): Synthesize the Val-13(S5)/Leu-17(S5) stapled Tesamorelin analogue directly. Measure helicity by CD spectroscopy, protease resistance by HPLC-based degradation assay (DPP-IV, chymotrypsin, plasma stability), and GHRHR binding by a competitive radioligand displacement assay. The hypothesis is strong enough on mechanistic grounds to warrant this without in silico validation.

Route B — Physics-based modelling: Apply molecular dynamics simulation with explicit CHARMM or AMBER force-field parameters for S5 residues and the olefinic cross-link. This would directly test the helicity pre-organization hypothesis in silico without relying on AF2's sequence-based prediction architecture.

Route C — Predictor-compatible Tesamorelin modifications: Pivot to modification chemistries within AF2's training distribution. Options that have not been tested on Tesamorelin in this lab include: C-terminal PEGylation for half-life extension (does not involve non-canonical backbone chemistry); conservative substitution of Ser-9 or Asn-8 with natural proteolysis-resistant residues; or fatty acid conjugation at an internal Lys residue analogous to the successful palmitoylation in Fold №48 (Ipamorelin, REFINED). These would generate interpretable predictions that could guide synthetic prioritization.

Route D — Shorter GRF fragment stapling: Test the i,i+4 hydrocarbon staple concept on a truncated GRF(1-17) fragment rather than the full 44-mer. Shorter stapled peptides are better represented in AF2's training data and may yield interpretable predictions, with the understanding that C-terminal truncation may affect GHRHR efficacy.