Phase 8h-lite #7 · Cochlear-compartment PBPK for v5.2__aq3__adamantyl__CONHOH__-Cl

Goal. Close the ELX-02 In Vitro Clinical PBPK Disconnect gap surfaced by DR1: RWM permeability alone (Phase 8h-lite #2) does not predict target-tissue exposure. Build an 8-compartment closed-form PBPK ODE model and quantify free-drug concentration at the hair-cell ER (the stereocilin folding compartment) as a function of time after intratympanic deposition.

Method. 8-compartment ODE: DEPOT (P407 gel) → MEC (middle-ear cleft) → PER_B (perilymph basal) ↔ PER_A (perilymph apical) ↔ ENDO (endolymph) → HC_CYT (hair-cell cytosol) ↔ HC_ER. Sinks: eustachian-tube clearance from MEC, stria-vascularis clearance from PER_B/PER_A/ENDO/HC_CYT to BLOOD. Rate constants from Salt & Ma 2001 (RWM permeability + ST t½), Salt & Plontke 2018 (longitudinal D), OTO-104/OTO-201 6% P407 (gel residence 42 h t½), and best-bracket estimates for hair-cell apical uptake (no published kinetic constant for hydroxamic-acid uptake into OHCs).

Script: scripts/phase8h_lite/cochlear_pbpk_ode.py. Run time <2 s. JSON: result_07_cochlear_pbpk.json.

Inputs

ParameterValueSource
Dose100 µM in 300 µL = 30 nmolreference applied conc per Phase 8h-lite #2
P_RWM1.45×10⁻⁷ cm/sPhase 8h-lite #2 (lead candidate, neutral fraction-corrected)
A_RWM1.164 mm²Salt & Ma 2001
V_DEPOT300 µL (gel reservoir)OTO-104 / OTO-201 IT injection volume
V_MEC100 µL (middle-ear free fluid)anatomical estimate
V_PER_B / V_PER_A35 / 35 µLhuman ST 70 µL split basal/apical
V_ENDO10 µLscala media, Hahn 2006 scaled to human
V_HC_CYT / V_HC_ER10⁻⁵ / 10⁻⁶ mL~10 000 OHCs × 1 pL × 10% ER fraction
t½ gel erosion42 hOTO-104 P407 6%
t½ MEC eustachian60 minnon-gelled vehicle clearance estimate
t½ perilymph → blood200 minSalt 2001 guinea-pig × volume scaling
t½ perilymph → endolymph400 min (0.5× blood)Sterkers 1988
D_long perilymph1×10⁻⁶ cm²/s over L 1.5 cmPlontke & Salt 2003
Kp ER:cyt1.0 (neutral acid, no trapping)Henderson-Hasselbalch logic
Target EC1 µM (Kd-implied)Phase 8h-lite #2 §interpretation

Outputs (3 apical-permeability scenarios)

Quantityfast (t½ 30 min)default (95 min)slow (300 min)
Cmax MEC6.52 µM6.52 µM6.52 µM
Cmax PER_B0.003 µM0.003 µM0.003 µM
Cmax PER_A0.00003 µM0.00003 µM0.00003 µM
Cmax ENDO0.001 µM0.001 µM0.002 µM
Cmax HC_CYT1.34 µM0.89 µM0.43 µM
Cmax HC_ER (target)1.34 µM0.90 µM1.18 µM
tmax HC_ER14.6 h12.0 h6.5 h
HC_ER @ 6 h0.77 µM0.44 µM0.14 µM
HC_ER @ 24 h1.21 µM0.81 µM0.40 µM
HC_ER @ 48 h0.81 µM0.55 µM0.27 µM
HC_ER @ 72 h0.55 µM0.37 µM0.18 µM
fraction of 72 h with HC_ER ≥ 1 µM39 %0 %0 %
mass balance @ 72 h100.00 %100.00 %100.00 %

Interpretation

The Phase 8h-lite #2 RWM permeability number was load-bearing only on the boundary itself. Once cascaded through the full cochlear compartment topology, the lead’s HC_ER exposure at the standard 100 µM applied IT dose:

  • Fails the 1 µM Kd-implied EC threshold in 2/3 scenarios. Default-apical gets 0% time above EC; slow-apical gets 0.2%. Only fast-apical (t½ 30 min lipophilicity-driven uptake) crosses ≥ 1 µM for 39% of the 72-h window.
  • The dominant loss is eustachian-tube clearance from MEC. At equilibrium between gel erosion (fastest delivery rate at t = 0) and eustachian tube (k = 1.16×10⁻² /min), MEC clears with t½ ≈ 60 min while RWM crossing has k = 1.0×10⁻⁶ /min — a 4-order-of-magnitude rate gap. Free drug in MEC is leaving via eustachian tube ≈ 11 000× faster than it crosses RWM. This is the structural reason RWM-permeability-only models over-predict cochlear exposure.
  • PER_A is essentially unreached. Cmax 3×10⁻⁵ µM (0.03 nM); the apical turn (speech-frequency processing region per Hair Cell Biophysical Reference Table) does not see therapeutic concentrations under any scenario. Longitudinal perilymph diffusion k = 2.7×10⁻⁵ /min (t½ ≈ 26 000 min ≈ 18 days) is the bottleneck.

Implications for H01

deliv-3 ceiling reinforced (held)

Phase 8g-v2 already pushed deliv 4 → 3 on APBS off-target failure. This PBPK result is an independent reason for the same ceiling, working from the delivery side rather than the selectivity side. Even if Phase 8g-v3 / Phase 8e-v2 close the off-target gap, deliv 3 holds until at least one of the following lifts cochlear exposure to a clinically tenable margin:

Three concrete improvement levers (P1-light each)

  1. Higher applied dose. 100 µM → 1 mM (10×). Linear scaling: HC_ER Cmax 8.9 µM (default), 13.4 µM (fast). 1 mM IT applied is well within reported ranges for steroid IT delivery (dexamethasone 4–24 mg/mL ≈ 10–60 mM) and DR5 hydroxamate-otoprotective SAHA precedents. Headroom is real. Lead Kd targeting 100 nM rather than 1 µM gets the exposure margin to 100×.
  2. Higher P_RWM via formulation. Salt 2018: inflammation increases P 2–10×. Co-formulating with a permeation enhancer (e.g., 0.5% sodium caprate or DMSO co-solvent) is a developability path that doesn’t require chemistry change. Conservative 3× P_RWM → HC_ER Cmax ≈ 2.7 µM (default).
  3. Higher Kd potency on the lead. v5.3 Tier-1 candidates (CONHOMe / phosphonate per h01 Phase 7I v52 Combined Boltz-2 Analysis 2026-04-26) target sub-µM Kd. If Phase 5d ensemble re-dock confirms 100 nM range, the 1 µM target EC drops to 100 nM and all three apical scenarios pass.

The three levers are independent and additive. Combined, they widen the therapeutic window by 30–100×.

One open mechanistic question PBPK does not close

The hair-cell apical permeability bracket (30–300 min t½) is the dominant uncertainty. Literature has no kinetic constant for hydroxamic-acid uptake across OHC apical membrane. This is a wet-lab gap (HEI-OC1 or UB-OC1 cell uptake assay), not a compute gap.

Caveats

  1. Apical t½ bracket is the soft underbelly. A 10× error in this single rate constant translates to 10× error in HC_ER exposure. Wet-lab grounding on this number is the highest-leverage refinement.
  2. No active transport. OHCs express uptake transporters (SLC22 family); their substrate scope for hydroxamic acids is unmapped.
  3. Single-bolus IT, no re-dosing. Multi-dose (e.g., q72h IT regimen, three injections over a week) was not modeled; would lift trough exposure significantly.
  4. No P407 rheology-coupled erosion kinetics. First-order release approximation; real P407 6% has zero-order plateau then accelerated dissolution near depletion. Refines tmax shape, not magnitudes.
  5. No protein binding. Plasma protein binding for hydroxamic acids varies 50–95%; not relevant for the cochlear free-drug compartment but a paper-§7 caveat.

Connections