FOLD №51 — TB-500 Thr-4 → 4-Fluoro-Phe

Verdict: DISCARDED | Peptide: TB-500 (LKKTETQ) | Class: REGENERATIVE Modified sequence: LKK(4F-Phe)ETQ | Target: G-actin, beta-actin (UniProt P60709)

Mechanism of action (background)

TB-500 (Ac-LKKTETQ) is the synthetic heptapeptide fragment corresponding to residues 17–23 of thymosin β4 (Tβ4), a ubiquitous 43-residue intracellular protein that sequesters G-actin monomers in a 1:1 complex, thereby regulating the pool of actin available for filament polymerization. The LKKTETQ sequence is the minimal pharmacophore responsible for Tβ4's characteristic biological activities: G-actin sequestration, promotion of cell migration, wound healing acceleration, and angiogenesis induction. In full-length Tβ4–actin co-crystal structures, the central LKKTET motif adopts a helical-turn conformation that lies along a shallow hydrophobic groove on G-actin subdomain 1, with the cationic LKK triad making electrostatic contacts with the acidic surface of actin and the central TTET segment contributing hydrophobic and polar contacts with the pocket residues Tyr-143 and Ile-341.

From a pharmacological standpoint, TB-500 promotes tissue repair, angiogenesis, and anti-inflammatory signaling by modulating the G-/F-actin equilibrium and downstream actin-dependent transcriptional programs (SRF/MRTF pathway). The peptide is short, largely unstructured in isolation, and undergoes rapid proteolytic degradation in vivo — PMID:38382158 demonstrates primary cleavage to Ac-LK within 0–6 hours, with the longer metabolite Ac-LKKTE (positions 1–5) detectable to 72 hours and retaining wound-healing activity in fibroblast assays.

This fold is the fifth distillation in the TB-500 series. Earlier work explored: N-terminal acetylation for metabolic protection (Fold №7, REFINED, pLDDT 0.87), Lys-2 → Orn substitution targeting the tryptic dibasic motif (Fold №16, DISCARDED), an i,i+3 Lys-3/Glu-6 lactam bridge for conformational preorganization (Fold №28, REFINED, pLDDT 0.81), and C-terminal palmitoylation for albumin-mediated half-life extension (Fold №38, REFINED, pLDDT 0.84). All prior REFINED folds targeted stability or PK. FOLD №51 is the first in this series to directly target binding affinity chemistry at the actin interface.

Modification hypothesis (what we tested)

The hypothesis was that replacing Thr-4 of the LKKTETQ heptapeptide with 4-fluoro-L-phenylalanine (4F-Phe) would convert a marginal polar contact into a hydrophobic aromatic anchor against the subdomain 1 cleft of G-actin. The rationale was threefold:

1. Geometric compatibility: Thr is a β-branched residue; 4F-Phe shares a similar Cβ volume and branching geometry, minimizing backbone distortion at the substitution site.
2. Hydrophobic pocket exploitation: If the Thr-4 hydroxyl points toward the apolar Tyr-143/Ile-341 face of actin rather than making a specific H-bond with a polar actin residue, replacing it with an aromatic ring would improve shape complementarity and hydrophobic packing energy.
3. Fluorine dipolar effects: The para-fluoro substituent adds a C–F···C=O dipolar contact and modulates ring quadrupole moment — a strategy well-documented in GPCR and protease SAR to enhance affinity 2–5× over unsubstituted Phe while suppressing metabolic aromatic hydroxylation.

This hypothesis is grounded in structural inference from full-length Tβ4–actin data and general fluorine medicinal chemistry principles. It is importantly distinct from prior TB-500 folds: all three REFINED predecessors (Folds №7, 28, 38) improved stability or half-life without addressing binding chemistry. FOLD №51 attempts to improve the interaction itself.

Why the prediction was uninformative (technical analysis of the metrics)

The structural predictor returned a well-folded peptide (pLDDT 0.83 is strong, consistent with Fold №28 at 0.81 and Fold №38 at 0.84 in the same series). The problem is the interface. The ipTM score of 0.48 means the model does not have sufficient confidence in the relative positioning of the peptide and the actin surface to produce a trustworthy binding pose. At this confidence level, we cannot determine whether the 4F-Phe ring is inserting into the subdomain 1 cleft as hypothesized, adopting an exposed rotamer away from the pocket, or simply not converging to a stable interface geometry. The Boltz-2 affinity module — which would have provided an independent predicted binding energy or probability estimate — returned no values, removing the second data stream. The absence of Chai-1 agreement data further means there is no cross-model corroboration.

Critically, this is not a case where the predictor returned a confident result pointing to poor affinity (which would be a genuine negative result). The predictor returned an uncertain result — it could not reliably model the complex at all. The ipTM failure likely reflects the intrinsic challenge of modelling a heptapeptide (only 7 residues) against a large 375-residue actin monomer, where the peptide represents a tiny fraction of total surface area and the signal-to-noise in complex prediction is inherently low. The 4F-Phe non-canonical amino acid may additionally fall outside the training distribution of the structural predictor in a way that degrades interface confidence beyond baseline.

The heuristic sequence-based profile (aggregation propensity 0.0, stability 0.85, half-life 15–45 min) is reassuring on the peptide-intrinsic properties — the substitution does not appear to introduce aggregation risk and the stability profile is favorable — but these metrics are not derived from the complex prediction and say nothing about binding.

What this tells us (negative results are data — what does it rule out?)

This distillation does not rule out the Thr-4 → 4F-Phe hypothesis. It rules out the current computational pipeline's ability to evaluate it reliably. The distinction matters:

• What is ruled out: That this modification can be scored by single-run Chai-1/Boltz-2 on a 7-residue peptide against a 375-residue actin monomer without a constrained template or structural prior. The ipTM threshold failure is a tools-limitation finding, not a chemistry finding.
• What is NOT ruled out: That 4F-Phe at position 4 improves G-actin binding affinity. The physicochemical rationale remains intact. The pocket geometry argument has not been tested — it has been untestable by this approach.
• What remains unknown: (1) Whether Thr-4 makes a productive H-bond or a marginal polar contact with actin — the entire hypothesis pivot point. (2) Whether 4F-Phe at this position alters protease recognition at the Ac-LKKTE cleavage site (PMID:38382158), potentially shortening or extending the in vivo half-life of the biologically active metabolite. (3) Whether the fluorinated aromatic improves or disrupts the helical-turn conformation of the central TTET segment that is required for actin engagement.

In the context of the TB-500 series, this fold establishes that binding affinity modification at the actin interface is currently outside the predictive reach of the single-run pipeline at this peptide size. Folds №7, 28, and 38 all succeeded in part because their modifications (acetylation, macrocyclization, lipidation) produced strong intramolecular structural signals that pLDDT captures well — conformational preorganization, capping, tail extension. Binding interface chemistry in a 7-mer requires a different approach.

Alternative hypotheses to test (avoid this failure mode)

1. Use the Fold №28 lactam-constrained scaffold as the base structure. Fold №28's i,i+3 Lys-3/Glu-6 lactam bridge (REFINED, pLDDT 0.81) preorganizes the helical-turn conformation. A double-modified peptide — lactam bridge + Thr-4 → 4F-Phe — would test whether conformational rigidity enables a more reliable interface prediction by reducing the entropy penalty of binding. The constrained scaffold may also improve ipTM by presenting a more defined interaction surface. This is the highest-priority next fold in this direction.

2. Expand the non-canonical aromatic series at position 4. If 4F-Phe is outside the predictor's training distribution, native phenylalanine (Phe) or 4-methyl-Phe at position 4 would test the hydrophobic anchor hypothesis with canonical or near-canonical residues that the predictor handles more reliably. A positive result with unsubstituted Phe would validate the aromatic pocket concept before committing to the fluorinated variant.

3. Test Thr-4 → Tyr as a step toward aromatic substitution. Tyrosine retains the Thr β-hydroxyl pharmacophore while adding an aromatic ring — a smaller hydrophobicity jump that could serve as a stepping-stone modification and might produce a more stable complex prediction. If Tyr is well-tolerated, the Thr → 4F-Phe step becomes more justified.

4. Ensemble docking with an explicit Tβ4–actin template. The fundamental limitation here is template availability. Using the published Tβ4–actin co-crystal structure (PDB entries from the Dominguez group) as a structural prior for constrained docking — rather than relying on de novo complex prediction — would provide far more reliable interface geometry for evaluating the 4F-Phe substitution. This is a computational strategy change, not just a sequence change.

5. Orthogonal wet-lab approach: fluorescence polarization binding assay. Given the complete absence of TB-500 Kd data in the published literature (a gap explicitly identified by the Literature agent), a simple FP binding assay with FITC-labeled TB-500 versus G-actin would establish a baseline affinity. Any chemical modification can then be benchmarked against this reference — making future in silico predictions interpretable in a quantitative context that currently does not exist.