DISTILLATION №66 — HUMANIN S7C/L11C i,i+4 Disulfide Staple

Verdict: PROMISING | Target: BAX (Q07812) | Class: LONGEVITY

Disclaimer: All findings are in silico predictions only. No wet-lab validation has been performed. Heuristic properties are sequence-derived estimates. This is not medical advice.

Mechanism of Action

Humanin (HN) is a 24-residue mitochondria-derived peptide that directly engages BAX (BCL-2-associated X protein), the central executor of the intrinsic apoptotic pathway. BAX resides in the cytosol in an inactive, monomeric conformation and, upon apoptotic signaling, undergoes conformational activation, oligomerization, and insertion into the mitochondrial outer membrane — triggering MOMP and cytochrome c release. HN physically binds BAX and sequesters it into fibrillar complexes that prevent this translocation (Guo et al., 2003; Morris et al., 2019). The central segment of HN (approximately residues 5–15) is the experimentally implicated BAX-binding pharmacophore, and HN's α-helical structure in this region is functionally load-bearing: mutations disrupting helical integrity abolish both anti-apoptotic activity and the characteristic fibrillar BAX-sequestration morphology. Beyond BAX, HN also engages BimEL, a BH3-only proapoptotic protein, through the same functional surface, indicating that the central helical segment must accommodate structurally distinct binding partners.

Performance Applications

Humanin analogs with enhanced BAX-binding affinity are of interest in the context of:

• Cytoprotection and longevity biology: HN expression declines with age and is inversely correlated with age-related cell loss in neuronal and cardiomyocyte populations. A conformationally stabilized, higher-affinity BAX inhibitor could extend the anti-apoptotic window in post-mitotic cells.
• Ischemia-reperfusion injury: The S14G HN analog (HNG) has demonstrated cardioprotective and renoprotective effects in preclinical I/R models, and a conformationally locked variant could offer improved potency in similar paradigms.
• Neurodegeneration research: HN was originally isolated from surviving neurons in Alzheimer's disease tissue, and BAX-mediated apoptosis is implicated in neurodegenerative cell loss. Research tools with tighter BAX engagement would be valuable for dissecting this mechanism.

Note: These are research contexts, not therapeutic indications. No clinical data exist for this specific variant.

Modification Rationale

The native Humanin sequence contains partial α-helical structure in solution, with the central segment (residues 5–15) being only incompletely organized in the free peptide — a conformational liability that limits affinity for BAX by imposing an entropic cost on folding-upon-binding. This fold introduces a disulfide staple between engineered Cys-7 and Cys-11, separated by exactly one α-helical turn (i,i+4 spacing), to pre-organize this pharmacophoric turn into its bioactive helical conformation before target encounter.

The i,i+4 geometry is the canonical spacing for α-helical turn stabilization via disulfide bridges: it enforces φ/ψ angles in the α-helical region across precisely one turn (~100° rotation, ~5.4 Å rise), maximizing the helix-nucleating geometric match. This is architecturally tighter and more directionally specific than the i,i+6 disulfide explored in Fold #22 (S14C/Cys-8 pair), which spanned a longer loop with weaker helical drive and received a PROMISING verdict at identical pLDDT (0.56). The step from i,i+6 to i,i+4 is the key structural refinement tested here.

Native Cys-8 is deliberately preserved as a free thiol. It sits at the i+1 position relative to engineered Cys-7, making it the most geometrically proximate potential mispairing partner. Selective oxidation of the 7–11 pair is assumed to be kinetically favored by the i,i+4 geometry, but this is an assumption that requires experimental validation (see Caveats).

Positions 7 and 11 were selected as surface-exposed residues on the predicted solvent face of the helix, not on the BAX-contact face. The native Ser-7 → Cys substitution removes a small polar residue; the Leu-11 → Cys substitution replaces a hydrophobic side chain with a smaller, polar one — the latter is the most significant chemical liability of the design and is discussed in Limitations.

Modified sequence: MAPRGFCCLLCLTSEIDLPVKRRA (Bold residues: engineered Cys-7 and Cys-11; native Cys-8 between them is preserved)

Predicted Properties — Where Signal is Moderate

Where the signal is meaningful: An ipTM of 0.58 for a 24-residue peptide against a globular protein places this prediction in the range where Boltz-2 demonstrates meaningful pose discrimination in benchmarks — the peptide is predicted to occupy a defined groove rather than making non-specific surface contact. This is consistent with the literature placing HN's central segment at the BAX interface.

Where the signal is insufficient: Per-residue pLDDT of 0.56 across residues 6–13 — the segment where the conformational hypothesis is most specific — means that the helical register and side-chain placement in the stapled region cannot be confirmed. The helix-tightening effect predicted by the i,i+4 geometry cannot be distinguished from the i,i+6 variant of Fold #22 at this resolution. The disulfide bond geometry itself is not verifiable from the pLDDT profile.

All heuristic values are sequence-based computational estimates, not experimental measurements.

What Would Strengthen This Signal

Additional computational predictions:

1. Chai-1 ensemble prediction with explicit disulfide bond encoding at positions 7–11: comparison of Chai-1 vs. Boltz-2 ipTM and pose RMSD would provide the most immediately actionable confidence upgrade. Convergent poses across two independent models would substantially strengthen the PROMISING verdict.
2. Free-energy perturbation (FEP) or MM-GBSA rescoring of the Boltz-2 pose to estimate ΔΔG relative to native HN and the Fold #22 variant — this would be the most direct computational test of whether the i,i+4 staple provides a binding affinity advantage.
3. MD simulation of the stapled peptide in isolation to confirm that the 7–11 disulfide drives φ/ψ angles into the α-helical region for residues 6–13, and does not introduce strain at Cys-8 (adjacent free thiol).
4. Disulfide selectivity modeling: Explicit computational assessment of the competing 7–8 (i,i+1) and 8–11 (i,i+3) mispairings would quantify the kinetic and thermodynamic favorability assumptions of the design.

Wet-lab experiments that would adjudicate this hypothesis:

1. Surface plasmon resonance (SPR) or ITC with purified recombinant BAX and the chemically synthesized S7C/L11C Humanin peptide (with 7–11 disulfide selectively oxidized by copper(II) catalysis or DMSO-mediated aerial oxidation in mildly acidic conditions to favor the i,i+4 pair): measure KD vs. native HN and the Fold #22 variant to test the affinity-enhancement hypothesis directly.
2. CD spectroscopy comparing the helical content of S7C/L11C-stapled HN vs. native HN vs. Fold #22 variant in oxidized and reduced conditions — this would directly test whether the i,i+4 staple increases α-helical content as predicted.
3. Mass spectrometry with differential alkylation of the three cysteines (7, 8, 11) to map disulfide connectivity in the oxidized product — confirming that the 7–11 pair forms preferentially over 7–8 or 8–11 mispairs.
4. Cell-based apoptosis protection assay (e.g., staurosporine-treated HEK293T or neuroblastoma cells, with TUNEL or caspase-3 readout) comparing the stapled variant vs. native HN vs. a reduced (DTT-treated) stapled control — this would test whether the disulfide constraint is functionally beneficial and whether cytosolic reduction abolishes the benefit.
5. Thioether analog synthesis (replacing the 7–11 disulfide with a non-reducible thioether or hydrocarbon staple) as a follow-up if the disulfide staple shows in vitro benefit but fails cell-based assays due to cytosolic reduction — directly testing the redox-liability hypothesis.

Lab Narrative & Cross-Fold Context

This fold continues the Humanin disulfide stapling series initiated in Fold #22, which introduced an i,i+6 Cys-8/S14C disulfide and received a PROMISING verdict at pLDDT 0.56 — strikingly identical confidence to the present fold. The architectural progression is meaningful: the i,i+4 geometry tested here provides a stronger stereochemical rationale for helix nucleation than the longer i,i+6 span, and positions 7 and 11 are predicted to be on the solvent face rather than the BAX-contact face, avoiding the occlusion risk present in Fold #22 where the S14C mutation was closer to the C-terminal end of the pharmacophoric segment. The fact that both disulfide variants converge on pLDDT ~0.56 and ipTM ~0.58 suggests this confidence ceiling may reflect the limitations of predicting a partially disordered peptide against BAX's dynamic surface rather than a failure specific to either design.

Fold #37 (S7A, DISCARDED, pLDDT 0.62) provides a useful control: removing Ser-7's hydroxyl without introducing a staple gave higher per-residue confidence but no productive interface, suggesting that position 7 tolerates substitution but that the staple geometry — not just the residue change — is doing chemical work here.

Fold #59 (N-terminal myristoylation, FAILED) reinforces that AlphaFold-family tools struggle with non-canonical chemical modifications, which is relevant context: the disulfide bond in the present fold is at least encodable within the standard amino-acid alphabet, giving it a structural prediction advantage over lipidated variants even if the confidence is modest.

The next logical step in this series — if Chai-1 corroboration supports the pose — would be a thioether-stapled analog at the same 7–11 positions (replacing the reduction-labile disulfide with a non-reducible covalent constraint), directly addressing the dominant cytosolic-redox liability identified by the literature agent.