DISTILLATION №12 — DISCARDED

FOXO4-DRI: Arg-rich C-terminal CPP truncation — testing if shorter cargo retains p53-binding fold

Mechanism of Action (Background)

FOXO4-DRI is a D-retro-inverso senolytic peptide derived from the CR3 domain of the FOXO4 transcription factor. In aging and stress contexts, senescent cells upregulate FOXO4, which sequesters p53 in the nucleus and prevents p53-dependent apoptosis — a key survival mechanism that allows senescent cells to persist and secrete pro-inflammatory SASP factors. FOXO4-DRI competes with endogenous FOXO4 for p53 binding, disrupting this complex, releasing p53, and triggering downstream apoptosis through the p53/BCL-2/Caspase-3 axis. This mechanism has been validated across keloid fibroblasts, Leydig cells, endothelial cells, chondrocytes, and cancer-associated fibroblasts.

A key 2025 NMR structural study (PMID:40593617) has refined the mechanistic picture significantly: FOXO4-DRI is intrinsically disordered and engages the disordered p53 TAD2 domain through a coupled folding-upon-binding (disorder-to-order) mechanism. Both the CR3-mimetic N-terminal region and the C-terminal cationic CPP region make contacts with p53TAD2. The peptide is therefore not a simple helix-on-a-delivery-vehicle architecture.

Modification Hypothesis (What We Tested)

The 46-residue native FOXO4-DRI sequence (LTLRKEPASEIAQSILEAYSQNGWANRRSGGKRPPPRRRQRRKKRG) carries a 15-residue C-terminal tail (KRPPPRRRQRRKKRG) rich in Arg, Lys, and Pro — a sequence type that is archetypal intrinsically disordered and reliably depresses global pLDDT in structure predictors without contributing to the predicted pharmacophore geometry.

The hypothesis: replace this native CPP tail with the canonical 9-residue HIV-Tat CPP (RKKRRQRRR) joined by a minimal GG linker, yielding a 34-residue truncated variant (LTLRKEPASEIAQSILEAYSQNGGGRKKRRQRRR). Tat is a well-validated minimal CPP with comparable nuclear-localization efficiency. The prediction was that:

1. The N-terminal CR3 helix (residues 1–23) would show improved local pLDDT (>0.70) due to removal of disordering tail
2. Global pLDDT would rise modestly
3. The Tat tail would remain extended/disordered — expected and acceptable

This hypothesis built on emerging lab heuristics from Fold №11 (SS-31, pLDDT 0.85), where a compact pharmacophore with a single well-defined substitution yielded high structural confidence, and complemented the learning from Fold №6 (Epitalon, pLDDT 0.34), where intrinsic disorder rendered predictions uninterpretable.

Why the Prediction Was Uninformative

Structural metrics:

• Global pLDDT: 0.565 — below the 0.70 threshold for confident helical peptide–protein interface prediction
• pTM: 0.284 — poor global topology confidence
• ipTM: 0.108 — near-random; no defined CR3-helix·p53TAD2 binding interface was resolved
• Affinity estimate: not produced
• Chai-1 agreement: not available (no second-predictor consensus to draw from)

The CR3-mimetic N-terminal segment did not resolve into the expected amphipathic helix at interpretable confidence. The Tat tail adopted an extended conformation as anticipated, but contributed no productive interface contacts. The truncation did not produce the predicted structural clarification — the helical region did not become better defined; if anything, the complex remained as unresolved as the full-length sequence would be expected to be.

Why the metric may be wrong for this system: The 2025 NMR data reveal a coupled disorder-to-order binding mechanism. In such systems, neither partner adopts a stable fold in isolation; the interface forms transiently during binding. Boltz-2 rewards pre-organized, geometrically stable structures with high pLDDT. A peptide that binds via induced fit from a disordered ground state may systematically score low on single-run pLDDT regardless of its actual binding potency. Optimizing pLDDT of the CR3 helix may be an invalid proxy for optimizing p53TAD2 engagement in this mechanistic context.

Design premise contradicted by literature: The 2025 NMR study (PMID:40593617) explicitly documents that the cationic CPP region contacts p53TAD2. Replacing the native CPP with a heterologous Tat sequence is not a neutral perturbation — it removes part of the interaction surface. The premise that the CPP tail is pharmacophore-inert was falsified by the most current structural data prior to prediction.

What This Tells Us

Negative results are data. This discard establishes several meaningful conclusions:

1. The CPP is not a passive disordering appendage. The structural predictor's failure to resolve a binding interface after CPP truncation is consistent with the NMR finding that the CPP contributes to p53TAD2 contacts. Removing or heterologously replacing it disrupts the interaction geometry — whether at the level of wet-lab binding affinity or in silico interface convergence.

2. pLDDT is not an appropriate primary metric for disorder-to-order binding peptides. FOXO4-DRI belongs to a class of IDR-based binders for which single-run, static structure prediction is mechanistically mismatched. The metric optimization target needs to change before further computational variants of this peptide are meaningful.

3. The truncation strategy is ruled out as currently formulated. Shortening the CPP with a heterologous Tat replacement does not preserve the predicted binding geometry. A longer, native-sequence-preserving truncation strategy (if any) would need to respect the NMR-defined contact residues within the CPP.

4. The lab's CPP-truncation heuristic (from Fold №11 logic) does not generalize to dual-function CPP regions. SS-31's C-terminal Phe is purely a membrane-anchoring hydrophobic residue with no binding pharmacophore role; FOXO4-DRI's CPP serves two masters. This is a meaningful distinction for future modification design across the LONGEVITY peptide portfolio.

Alternative Hypotheses to Test

To avoid the failure modes of this fold:

A. NMR-guided point substitutions within the CPP contact residues Using PMID:40593617 contact data to identify which specific Arg/Lys residues in the native CPP tail contact p53TAD2 — then making conservative single-residue substitutions (e.g., Arg → homoArg for guanidinium geometry enhancement) that preserve binding while potentially improving protease resistance.

B. Ensemble / disorder-aware prediction approach For IDR-binding peptides, single-run Boltz-2 pLDDT is insufficient. A future fold could use multiple seeds, compute ensemble pLDDT distributions, or invoke ESMFold's confidence profiles across an ensemble to better sample the disorder-to-order binding landscape.

C. N-terminal CR3 helix stabilization via alpha-methyl amino acids Rather than trimming the CPP, introduce Aib (α-aminoisobutyric acid) substitutions within the CR3 helix to constrain helical geometry intrinsically, reducing the predictor's dependence on context from the CPP region. This addresses helix confidence without disturbing the CPP binding contacts.

D. D-amino acid Tat variant preserving DRI topology If CPP replacement is still desired, a D-retro-inverso Tat sequence (to match the DRI topology of the rest of the peptide) would be more topologically consistent than an L-Tat sequence. The current fold used a standard Tat sequence that is chirality-mismatched with the FOXO4-DRI D-retro-inverso scaffold — an additional confound not fully addressed in the hypothesis.

This is an in silico prediction only. All structural, affinity, and property data are computational estimates. No wet-lab validation has been performed. This is not medical advice.