STRC Research

STRC Engineered Homodimer Avidity

7 min read

STRC Engineered Homodimer Avidity

STATUS 2026-04-23 — post-audit + post-Phase-1 + post-ipSAE:

  • Phase 1 AF3 FAILED: all 4 single-point mutants (R1581Y/F, S1579W/F) destabilised the homodimer vs WT (mean homodimer ipTM 0.15-0.19 vs WT 0.28-0.30). R-R repulsion hypothesis falsified on structural grounds — see STRC Engineered Homodimer Phase 1 Results.
  • ipSAE 0.000 (Dunbrack 2025): reassessing Ultra-Mini homodimer AF3 model_0 with the Dunbrack ipSAE metric — which corrects ipTM’s known disordered-region false-positive failure mode — gives zero interface residues passing PAE<10 Å cutoff. The prior ipTM 0.28-0.30 “uptick vs baseline” signal is now identified as exactly the kind of inflation ipSAE was designed to catch. Mech 3→2. Structural symmetry (94% C2) + convergent self-contacts at aa 1579-1581 survive (observations independent of the confidence metric). See STRC ipSAE Cross-Complex Reassessment 2026-04-23.
  • Kd baseline unmeasured: STRC × TMEM145 monomer Kd has NO SPR/BLI/ITC measurement (see strc-tmem145-interactions + STRC TMEM145 Kd Experiment Plan). The “100 nM” figure below is derived from AF3 ipTM 0.43 — ipTM is NOT a Kd proxy (R² ≈ 0.06 for protein-protein; see avidity-and-dimers §3).
  • Any quantitative avidity claim (100×, 1000×) is therefore a dimensionless ratio applied to an unmeasured baseline. The avidity principle is textbook-valid (Jencks 1981 PNAS; Mammen et al. 1998 Angew Chem; Kramer & Karpen 1998 Nature). The specific Kd numbers in this note are placeholder until experimental measurement.

Hypothesis #26. Single point mutation on the Ultra-Mini homodimer interface converts a weak self-association (AF3 ipTM 0.28-0.30) into a stable dimer, giving Mini-STRC two TMEM145-binding surfaces per molecule. ⚠ The “100-1000× effective Kd drop” claim is a ratio applied to an unmeasured STRC×TMEM145 baseline — structurally motivated, quantitatively unverifiable until SPR/BLI.

The idea in one paragraph

STRC Homodimer Interface From CIF showed that Ultra-Mini retains a weak C2-symmetric self-contact at ARM 1579-1581 (all 5 AF3 models, 94% C2, 98% solvent exposure). STRC Ultra-Mini Full-Length TMEM145 AF3 showed Ultra-Mini × TMEM145 ipTM 0.43 — on-target but not confident. Standard biophysics: if a weak-Kd monomer dimerizes, the dimer × ligand apparent Kd collapses by Kd_dimer / [monomer] (avidity-and-dimers §1: Jencks-Mammen chelate formula). The specific numbers below (100 nM → 1 nM) are placeholder: the 100 nM input is not measured (ipTM cannot be converted to Kd); the 1 nM output is therefore arithmetic on a phantom input. What would remain valid post-measurement: the dimensionless avidity ratio itself, given C_eff typically 0.1-10 mM for tethered interfaces (avidity-and-dimers §1).

Strategic value (not contingent on exact Kd): if a stabilised dimer confirms in wet-lab SEC/AUC, it delivers avidity through the same clinical construct (Anc80L65 + B8-IgK-Ultra-Mini-WPRE3-bGH, same CpG-depleted CDS) at zero delivery-axis cost. The gamble: one aa change vs re-architecting payload.

Mutation candidates

Contact residues identified in STRC Homodimer Interface From CIF: aa 1579, 1580, 1581 on both chains, homotypic self-contacts in all 5 AF3 models (A.1579↔B.1579, A.1580↔B.1580, A.1581↔B.1581). Actual residues at these positions (verified against UniProt Q7RTU9):

  • aa 1579 = S (serine)
  • aa 1580 = G (glycine)
  • aa 1581 = R (arginine)
  • aa 1582 = H (histidine), context ...SFLRQSGRHVSHL...

Why the dimer is weak: R-R repulsion. Two arginine side chains across a C2 interface (A.R1581 ↔ B.R1581) are classically destabilising — like-charged residues repel unless compensated by specific H-bond geometry. This is the most plausible reason AF3 confidence stays at ipTM 0.28-0.30 despite the geometric evidence (94% C2 symmetry, 98% surface exposure). The interface WANTS to form but the charge pair opposes.

Candidates for single-point stabilisation:

  • R1581Y — replaces positive charge with aromatic tyrosine; tyrosine-tyrosine across C2 provides π-π stacking AND potential hydrogen bond via OH. Removes repulsion, adds favourable contact. Most promising mechanistically.
  • R1581F — replaces positive charge with phenylalanine; pure aromatic stack, no OH to worry about, shorter-range but less entropy cost.
  • S1579W — tryptophan pi-stack across the C2 axis at the adjacent contact residue (leucine-zipper precedent, Alber 1992). Neutral→neutral substitution, pure aromatic-stacking gain.
  • S1579F — phenylalanine symmetric stack at position 1579; less steric risk than W.

Phase 1 AF3 tests the four single mutants + one WT reference; if any crosses homodimer ipTM ≥ 0.50 AND preserves Ultra-Mini × TMEM145 ipTM ≥ 0.40 per subunit, it’s a winner. R1581Y and R1581F are the most mechanistically grounded bets; S1579W/F are the safety-net neutral-to-aromatic options.

Phase 1 AF3 batch prepared (2026-04-23)

Builder ~/STRC/models/af3_jobs_2026-04-23d_engineered_homodimer_builder.py generates 5 AF3 jobs in ~/STRC/models/af3_jobs_2026-04-23d_engineered_homodimer/:

  1. engineered_homodimer_WT_ref — wild-type Ultra-Mini × 2, TMEM145 × 2 (baseline reference)
  2. engineered_homodimer_S1579W
  3. engineered_homodimer_R1581Y
  4. engineered_homodimer_S1579F
  5. engineered_homodimer_R1581F

Each is a 4-chain AF3 complex: 2 × Ultra-Mini (mutant) + 2 × TMEM145 (full 493 aa). Total 4 × ~600 aa = 2388 aa per job. Seed 42. Gate per job: chain-pair A↔B ≥ 0.50 (dimer stabilised) AND chain-pair A↔C / B↔D ≥ 0.40 (each subunit preserves TMEM145).

Awaiting user submission to AlphaFold Server (5 jobs, daily limit 20, 16/20 used today = submit with next daily quota).

Why this beats SpyCatcher (#10) and TECTA chimera (#11)

Both of those added EXTRA protein content (SpyCatcher 130 aa, TECTA 500 aa) to get multivalency or anchoring. Result: architectural tax at the binding interface (SpyCatcher costs 0.05-0.08 ipTM; TECTA occludes). This hypothesis adds zero new content — just flips one existing residue. No new AAV payload. No new immunogenicity concern (WT STRC has paralog variants with W at this position in mouse, suggesting tolerance).

Why this beats #4 Strategy B mRNA-LNP and #9 Hydrogel as avidity source

Strategy B and Hydrogel try to compensate for weak binding via dose (mg-scale topical / systemic load). Engineered homodimer compensates via intrinsic affinity — ratio only, since absolute Kd is unmeasured. At cochlear PK — where dose is always dose-limited by round-window surgery trauma — any intrinsic Kd improvement is worth more than a dose-ceiling nudge. The “100-1000×” magnitude is the Jencks-Mammen-Kramer upper bound for well-tethered dimers (C_eff 1 mM, Kd_mono 100 nM → 10,000× ceiling); real-world ratios are linker-geometry sensitive.

Phase 1 plan

Single AF3 batch:

  1. Ultra-Mini(S1579W) homodimer × 2 TMEM145 — does W-W stack stabilise and preserve TMEM145 × 2?
  2. Ultra-Mini(R1581Y) homodimer × 2 TMEM145
  3. Ultra-Mini(S1579F) homodimer × 2 TMEM145
  4. Ultra-Mini(R1581F) homodimer × 2 TMEM145

Gate per candidate: homodimer ipTM ≥ 0.50 (up from 0.28-0.30 wild-type) AND each Ultra-Mini × TMEM145 ipTM ≥ 0.40 (monomer baseline).

If any passes → Phase 2: wet-lab expression + SEC/AUC to confirm dimer in solution at physiological concentration. ~$10-20k, 8-12 weeks. If SEC-AUC confirms → Phase 3: HEK293 co-IP (STRC-mut × TMEM145) side-by-side with WT Ultra-Mini — must show 10-100× shift in apparent Kd.

Risks

  1. Stabilised dimer may aggregate. If interface becomes too sticky, Ultra-Mini could polymerize into insoluble chains. SEC/AUC gate catches this.
  2. Dimer may occlude its own TMEM145 interface. Aa 1579-1581 are in the ARM repeat region (1075-1775 Ultra-Mini numbering, internal ARM position); TMEM145 contacts in 1603-1749 GOLD zone are ~100 residues away, but tertiary proximity possible. AF3 dimer × 2 TMEM145 job tests this.
  3. Mouse paralog tolerance is an assumption. Need to BLAST human STRC S1579 in mouse/primate orthologs to confirm the position tolerates aromatic residues naturally.
  4. Homodimer might be AF3 artifact after all. If wet-lab SEC-AUC shows Ultra-Mini is strictly monomeric at physiological concentration, the whole premise collapses. Same risk was already flagged in STRC Homodimer Interface From CIF as “weak-positive pending wet-lab.”

Scoring (post-Phase-1 fail + post-audit 2026-04-23)

  • Mechanism 3/5 (was 4): avidity principle is textbook-valid (Jencks 1981, Mammen 1998, Kramer & Karpen 1998), but Phase 1 AF3 ALL FAIL — single-point R/S→Y/F/W on the 1579-1581 interface destabilises the homodimer; R-R repulsion premise contradicted. Any forward step now requires multi-point or out-of-interface interventions.
  • Delivery 5/5: unchanged — same chassis as S-tier #3.
  • Misha-fit 4/5: unchanged — works on paternal null allele.

min(3, 5, 4) = 3 → B-tier (was A on arrival). Blocked at two independent gates: (a) Phase 1 AF3 gave no viable single-point variant; (b) STRC×TMEM145 Kd baseline unmeasured. Neither can be fixed by compute alone. Promotion back requires either a wet-lab Kd measurement (Plan A/B in STRC TMEM145 Kd Experiment Plan) or a new structural hypothesis for the interface (e.g., disulfide engineering per avidity-and-dimers §4 benchmarks).

Connections

Graph View

Backlinks