STRC Research

Phase 8h-lite #8 · K1141 pocket vs stereocilin-partner interface map

5 min read

Phase 8h-lite #8 · K1141 pocket placement on the E1659A AF3+MD model

Goal. Address the Migalastat Dosing-Cycle Principle design constraint for an extracellular pharmacochaperone: if K1141 pocket binding overlaps a published stereocilin-partner contact patch, a chaperone bound there could disrupt tectorial-membrane assembly even after successful trafficking. Compute the structural substrate for that overlay using the existing Phase 5d full-length E1659A snapshot.

Method. Manual PDB column parser on models/artifacts/phase5d_snapshots/snap_010.pdb (1775-residue full-length E1659A, single MD frame from Phase 5d 2 ns trajectory). Pocket-shell residues at 6 / 8 / 12 Å of K1141 NZ. Per-residue heavy-atom-burial proxy (freesasa not installed; quartile-based SASA ranking instead — relative ordering valid, absolute Ų indicative only). Domain assignment from UniProt Q7RTU9 / SMART / Pfam published architecture.

Script: scripts/phase8h_lite/k1141_interface_map.py. Run time <2 s. JSON: result_08_k1141_interface_map.json.

Findings

1. K1141 and E1659A are in different domains

ResidueDomainSASA-proxyBurial state
K1141central FN3-like (≈ 800–1200)120partially buried — pocket apex
A1659 (E→A mutation)C-terminal ZP-like (≈ 1500–1771)60fully buried — domain core

Inter-residue distance (K1141 NZ → A1659 CA): 13.91 Å.

The mutation site and the binding pocket sit in distinct structural regions ~14 Å apart. This rules out a “the pocket is the mutation site” reading of mech 4 — the lead does not bind to the mutated residue; it binds to a neighbouring electropositive pocket whose physics is altered by the mutation’s destabilising effect on the broader fold.

2. K1141 pocket is enclosed, not partner-facing

Pocket-shell composition (47 residues within 12 Å of K1141 NZ):

QuartileCountSASA range
Top quartile (most exposed, surface-facing)13≥ 133
Middle two quartiles (ambiguous)2280–133
Bottom quartile (buried, core)12≤ 80

K1141 itself sits at SASA-proxy 120 — in the middle band. This is the canonical Halgren druggable-pocket signature: not surface-exposed, not deep-core, but enclosed-yet-reachable. Phase 8h-lite #1 already characterised the pocket as druggable (V 1145 ų, phobic 0.61, V_pocket/V_lig 2.75); this result adds the structural-context layer.

3. Surface runs adjacent to K1141 are short (≤ 3 contiguous residues)

Surface-exposed pocket-shell residue runs in the FN3-like domain (gap ≤ 2 residues):

  • 1103 (1 res), 1121–1122 (2 res), 1125–1128 (3 res), 1144 (1 res), 1173 (1 res)

No run ≥ 5 contiguous surface-exposed residues — the published threshold for a continuous partner-binding interface patch (Jones & Thornton 1996; Lo Conte 1999). The K1141 region of the FN3-like domain has small isolated solvent-exposed faces interspersed with buried core, not a partner-recognition surface.

4. A1659 (mutation site) sits adjacent to longer surface runs in the ZP-like domain

Surface runs near A1659 (12-Å pocket of K1141 NZ extends into ZP-like domain):

  • 1612 (1 res), 1619 (1 res), 1643 (1 res), 1648 (1 res), 1651 (1 res)

These are also short isolated runs at this snapshot, but the un-mutated version of the ZP-like domain hosts the canonical tectorial-membrane assembly interface — α/β-tectorin (TECTA/TECTB), CEACAM16, OTOG, OTOGL contacts per Verpy 2008 / Kammerer 2012 / Lukashkin 2012 (literature attestation pending lit-mining task; cite confirmation before paper-quoting). The fact that the destabilised mutant snapshot has fragmented its ZP-like surface is itself consistent with the mech-4 story: E1659A perturbs the ZP-like fold, including its presentation surface to partners.

Direct mechanistic implications for h01

Migalastat-design-constraint check: passes

Per Migalastat Dosing-Cycle Principle, an extracellular target without a pH-gradient dissociation site requires either tuned intrinsic k_off OR allosteric binding outside any stereocilin-partner interface. Result #2 + #3 above support the second branch: K1141 pocket is enclosed within FN3-like domain core, has no continuous surface-exposed run ≥ 5 residues nearby, and is in a different domain from the canonical partner-binding ZP-like region. A chaperone bound at K1141 is unlikely to disrupt tectorial-membrane assembly post-trafficking.

This shifts the residence-time design from “must dissociate at the right place” (Migalastat case) toward “may safely persist on the protein” (Tafamidis case) — the chaperone can in principle stay bound through trafficking, ECM crosslinking, and into mature stereocilin tenure without sterically blocking partner contacts. Slow k_off becomes a pure asset, not a double-edge.

Cross-link to CFTR F508del + Elexacaftor precedent

Per DR1 — Pharmacochaperone Clinical Precedents §5, Elexacaftor binds an allosteric hydrophobic pocket on CFTR distinct from the F508del mutation site (NBD1 vs the corrector-binding TMD1/TMD2 interface). Stereocilin’s geometry (K1141 in FN3-like core, E1659A in ZP-like core, ~14 Å apart) is the direct architectural analog of CFTR’s geometry (corrector pocket ≠ F508 site). The lead K1141 binding strategy inherits Elexacaftor’s regulatory and mechanistic playbook.

One claim this result does not yet close

The “K1141 ≠ partner-interface” argument here is inferred from local structural features of a single MD frame, not from a primary-literature stereocilin interactome. The lit-mining companion task (tracked in next_step (b) per h01 log DR1-distillation entry) is the literature-side closure: confirm by published interactome that no stereocilin-CEACAM16 / -OTOG / -OTOGL / -OTOA / -TECTA / -TECTB contact patch maps onto FN3-like residues 1100–1175 (the K1141-pocket sequence neighbourhood). Until that closure lands, this result is structural evidence, not a literature-grounded interactome claim.

Caveats

  1. freesasa not installed — heavy-atom-burial proxy used; quartile ranking valid relatively, absolute Ų indicative only. Re-run with pip install freesasa for publication numbers.
  2. Single MD frame. snap_010 from Phase 5d 2-ns trajectory; surface accessibility fluctuates over MD. Better practice: average SASA across all 20 Phase 5d snapshots. Held back as P1-light refinement; this single-frame check captures the qualitative geometry.
  3. Domain boundaries from public annotation. UniProt Q7RTU9 SMART/Pfam domain ranges are predictive; experimental boundaries may differ ± 30 residues. The FN3-like ≠ ZP-like split at residue ~1200 is robust on the architecture-level argument made here.
  4. No homodimer interface considered. h26-engineered-homodimer hypothesis explores STRC homodimerisation; if K1141 sits at a homodimer interface in the engineered-dimer geometry, bivalent binding could stabilise the monomer-monomer contact — orthogonal mechanism, not addressed here.

Connections

Graph View