FOLD №65 — TB-500 Lys-3 → D-Lys: Chirality Inversion for Protease Resistance

Verdict: DISCARDED | Class: Regenerative | Target: β-actin (P60709)

TLDR

Fold №65 was DISCARDED due to a tool-limit failure: the predicted protein–peptide complex returned an ipTM of 0.47, below the minimum confidence threshold required to trust the interface geometry. This is not a biological invalidation of the D-Lys-3 hypothesis — it is a failure of current structure prediction infrastructure to model single D-amino acid substitutions in short heptapeptides with sufficient resolution. AlphaFold-class models, including Boltz-2, have no explicit stereochemical representation of D-amino acids; the prediction almost certainly treated D-Lys as L-Lys, rendering the output uninformative for the specific question asked.

What We Tried

TB-500 (Ac-LKKTETQ) is the minimal bioactive heptapeptide of thymosin β4, with the LKKT N-terminal motif responsible for G-actin sequestration and downstream regenerative signalling. Its primary metabolic liability is well-documented: Rahaman et al. (2024) established that serine protease cleavage at or adjacent to the dibasic K2–K3 stretch generates Ac-LK as the dominant early plasma metabolite — a biologically inactive two-residue fragment. TB-500's regenerative potential is therefore substantially curtailed by rapid proteolytic degradation.

This fold hypothesized that inverting the stereochemistry of Lys-3 to D-configuration would abolish productive trypsin/plasmin cleavage at the K2–K3 scissile bond, exploiting the near-absolute L-stereospecificity of serine protease S1 pockets at the P1 position. Crucially, the modification preserves full side-chain length and charge at position 3 (D-Lys has the same ε-amine and the same basic character as L-Lys), and leaves Lys-2 in its native L-configuration to maintain the canonical LKKT actin-binding pharmacophore. This directly addresses the lesson from Fold №16 (Lys-2 → Ornithine, DISCARDED), where shortening the Lys-2 side chain disrupted actin contact — here, no side-chain atoms are removed or altered, only the Cα chirality at position 3 is changed.

Why It Was Discarded

The primary discard reason is insufficient interface confidence: ipTM = 0.47 at the TB-500/β-actin complex level does not provide actionable structural information about whether the D-Lys-3 modification preserves or disrupts the predicted binding pose. The peptide-level pLDDT of 0.82 is adequate, but this metric reflects intrinsic peptide fold quality, not complex geometry — and for a 7-residue peptide docked against a large globular protein, interface confidence (ipTM) is the decisive metric.

A compounding and arguably more fundamental limitation is that Boltz-2 (and AlphaFold-class models generally) lack explicit D-amino acid stereochemistry in their training data and energy representations. The model almost certainly predicted the structure as if D-Lys were L-Lys, meaning the output does not represent the actual modified peptide at all. The Boltz-2 affinity module returned no values, and no Chai-1 ensemble was generated for cross-validation — removing the two independent checks that would ordinarily allow us to assess agreement. This pattern matches Fold №51 (Thr-4 → 4F-Phe, DISCARDED, pLDDT 0.83), where non-canonical substitution at a central position also failed to produce a trustworthy interface prediction.

What This Doesn't Mean

DISCARDED does not mean disproved. The biochemical rationale for D-Lys-3 is mechanistically sound and grounded in well-established serine protease stereochemistry — a principle validated across many therapeutic peptide programmes, even if not yet tested in TB-500 specifically. The prediction infrastructure simply cannot evaluate this specific modification: D-amino acids are invisible to current AlphaFold-class models, and the heptapeptide/actin interface is at the resolution limit for confident complex modelling with a single prediction run. The hypothesis that D-Lys-3 abolishes K2–K3 cleavage while preserving LKKT actin engagement is scientifically coherent, novel, and experimentally testable — it has simply not been adjudicated by this fold's tools. The discard reflects a measurement gap, not a negative result.

What Would Answer the Question

• Protease resistance assay (wet lab): Incubate Ac-LK[D-K]TETQ with human plasma, recombinant trypsin, or plasmin at physiological concentrations; monitor metabolite appearance by LC-MS/MS against the Rahaman et al. (2024) metabolite panel (Ac-LK, Ac-LKK, Ac-LKKTE). Comparison with native Ac-LKKTETQ and the Fold №28 lactam variant would create a systematic stability dataset for TB-500 analogues.
• G-actin binding assay (SPR or ITC): Surface plasmon resonance or isothermal titration calorimetry with monomeric G-actin and Ac-LK[D-K]TETQ to directly measure Kd relative to native TB-500. This would resolve the key unknown — whether D-Lys at position 3 is tolerated by the actin binding interface.
• Molecular dynamics / FEP with D-amino acid-aware force fields: Classical MD using CHARMM36m or AMBER ff19SB with explicit D-Lys parameters can correctly represent the inverted Cα geometry and sample the actin-binding interface conformation. Free energy perturbation (FEP) would provide a quantitative ΔΔG estimate for L→D chirality at position 3. This is the most appropriate computational approach given that AlphaFold-class tools cannot address this question.
• Combination with Fold №38 (C-terminal palmitoylation): If D-Lys-3 is confirmed to block K2–K3 cleavage, combining it with the REFINED C-terminal palmitoylation strategy from Fold №38 (Lys-7/γGlu-Palm) could address both protease-mediated degradation and albumin-mediated half-life extension simultaneously — a dual-mechanism stability analogue worth designing once each component is experimentally validated.

Raw Metrics

*Heuristic sequence-based estimates only — not wet-lab measurements and not specific to the D-Lys-3 modification.