FOLD №77 — TB-500 N-terminal Myristoylation

Verdict: DISCARDED (Tool-Limit Failure)

TLDR

Fold #77 was DISCARDED due to a tool-limit failure: current structure predictors (AlphaFold/Boltz-2/Chai-1) cannot model the N-terminal myristoyl (C14 fatty acid) modification, meaning the structural output reflects the bare LKKTETQ backbone only — not the actual Myr-LKKTETQ molecule under investigation. The Boltz-2 affinity module returned no binding probability values, and the ipTM of 0.60 is insufficient for a confident complex-level verdict. This is not a biological invalidation of the membrane depot hypothesis.

What We Tried

This fold tested N-terminal myristoylation of TB-500 — specifically, covalent attachment of a myristoyl (C14 saturated fatty acid) chain to the α-amine of Leu-1 via a stable amide bond, yielding Myr-LKKTETQ. The hypothesis was that the C14 chain would partition into plasma membranes, creating a reversible lipid depot from which the active heptapeptide would gradually release into the cytosol to engage its target: G-actin sequestration via the canonical LKKT pharmacophore.

This is the lab's second lipidation strategy for TB-500 and the first to target the N-terminus. Fold #38 established that C-terminal palmitoyl-γGlu lipidation (LKKTETK(γGlu-Palm)-NH₂) was REFINED with pLDDT 0.84, validating lipidation as a productive direction for this peptide. Fold #77 was designed as a complementary, mechanistically distinct approach: where Fold #38 targeted albumin-mediated systemic half-life extension via a C16 chain at the C-terminus, this fold targeted membrane insertion and intracellular depot formation via a shorter C14 chain at the N-terminus — a strategy known from endogenous myristoylated proteins (Src, HIV-1 Gag, MARCKS).

The N-terminal position was also chosen because aminopeptidase cleavage at Leu-1 is a documented TB-500 liability (Rahaman et al., 2024, showing Ac-LK as the primary short-interval metabolite). The existing commercial formulation uses acetylation (Ac-LKKTETQ) to partially address this; a myristoyl amide bond was hypothesized to provide more robust protection. This builds directly on the rationale established in Fold #7 (N-terminal acetylation) and addresses a gap that the acetyl cap only partially closes.

Why It Was Discarded

The discard is a tool-limit failure, not a structural or biological red flag:

• AlphaFold2/Boltz-2 blind to lipid chemistry: Standard structure-prediction pipelines process proteinogenic amino acid sequences. The myristoyl chain — a non-canonical chemical modification attached to the Leu-1 α-amine — is stripped or ignored by the model. The resulting prediction describes the LKKTETQ backbone geometry, not Myr-LKKTETQ behavior in a membrane environment.
• No affinity module output: The Boltz-2 binder probability module returned no values — the most direct signal that the modified molecule was outside the model's evaluable chemical space.
• ipTM 0.60: Complex-level confidence fell below the threshold for a reliable binding call. On a heptapeptide without reliable interface atom placement for the modified terminus, this is expected rather than informative.
• High pLDDT (0.87) is an artifact: The strong backbone confidence score reflects the model's familiarity with LKKTETQ as a short, well-behaved polar sequence — not confidence in the myristoylated molecule. A high pLDDT on a stripped sequence should not be interpreted as structural validation of the modification.
• Chai-1 agreement unavailable: No ensemble cross-check was possible to assess prediction stability.

Biological concerns (not the primary discard reason, but relevant context):

• Leu-1 and Lys-2 are directly adjacent to the actin-binding pharmacophore; a C14 chain at this position may impose steric costs on LKKT–actin engagement that acetylation does not.
• C14 chains on heptapeptides may drive micelle formation at physiological concentrations (CMC considerations), rather than monomeric membrane insertion.
• No published data exist on any lipidated LKKTETQ analog; the membrane depot mechanism is entirely de novo.

What This Doesn't Mean

DISCARDED does not mean disproved. This verdict reflects the boundaries of the current in silico toolchain, not a judgment on the biological validity of N-terminal myristoylation for TB-500. The membrane depot hypothesis is pharmacologically coherent, literature-supported in analogous systems (Src-family kinases, lipidated cell-penetrating peptides), and mechanistically distinct from anything the lab has tested before for this peptide. The failure to obtain a binding probability value means the question was never answered — not that the answer is negative. It is entirely possible that Myr-LKKTETQ inserts into membranes with appropriate kinetics, releases active peptide into the cytosol, and engages G-actin with preserved or enhanced potency. That question requires wet-lab or specialized computational tools to resolve.

What Would Answer the Question

• Solid-phase peptide synthesis + HPLC purification of Myr-LKKTETQ, followed by SPR (surface plasmon resonance) against immobilized G-actin (rabbit muscle, Cytoskeleton Inc.) to directly measure K_D relative to Ac-LKKTETQ. This is the gold-standard binding assay for this target class and is directly analogous to the approach used in Tβ4 fragment SAR studies.
• CMC measurement and liposome partitioning assay (pyrene fluorescence probe or DLS) to determine whether Myr-LKKTETQ forms micelles at physiologically relevant concentrations (1–100 µM range) or partitions as monomer into model POPC/POPE membranes — directly testing the depot hypothesis.
• Coarse-grained MD simulation (MARTINI force field, GROMACS/CHARMM-GUI) of Myr-LKKTETQ in an explicit POPC bilayer to model membrane insertion depth, orientation, and flip-flop kinetics — the computational tool class appropriate for lipid-modified peptides that AF2 cannot handle.
• Aminopeptidase protection assay (leucine aminopeptidase, human serum stability) comparing Myr-LKKTETQ vs. Ac-LKKTETQ degradation kinetics, directly testing the N-terminal cap hypothesis from the Rahaman et al. (2024) metabolic vulnerability data.

Raw Metrics

All values are in silico predictions. No wet-lab validation has been performed. This report is research context only, not medical advice.