# Preclinical & Translational Workflow Roadmap
## pMHC-II Multimer Allergen-Specific Immunotherapy for Eosinophilic Esophagitis

**Version:** 1.0
**Lead mechanism:** pMHC-II multimer-induced anergy (Mechanism 1; see pmhc_tcr_mechanism_review.md)
**Companion documents:** allergen_epitope_panel.csv (Task 1), recruitment_stratification_rationale.md (Task 2), t_cell_assay_protocol.md (Task 3)

---

## Overview

This roadmap sequences the work from validated computational epitopes to an IND filing and first-in-human study, in four phases. Each phase lists inputs, activities, decision gates (go/no-go), deliverables, timeline, and quantitative success criteria. Timelines are illustrative planning estimates for a well-resourced program.

```
Phase 1 (0-9 mo)     Phase 2 (6-18 mo)      Phase 3 (15-30 mo)     Phase 4 (30-42 mo)
Ex vivo human   ->   Reagent + in vivo  ->  IND-enabling GLP   ->  Phase 1b/2a
T-cell validation    PoC in models          tox/PK/CMC             all-comers trial
     |                    |                       |                      |
   GATE A               GATE B                  GATE C                 readout
(target engaged?)   (in vivo efficacy?)    (safe + manufacturable?)  (clinical signal?)
```

---

## PHASE 1 — Ex Vivo Human T-Cell Validation (Months 0–9)

**Goal:** Prove that dominant food-allergen epitopes drive genuine allergen-specific CD4+ responses in EoE patients, and that pMHC-II reagents engage and silence those T cells.

**Inputs:** Task 1 epitope panel; Task 2 per-patient personalized epitope prediction (HLA-agnostic — every patient's own genotype yields their panel); Task 3 assay SOP.

**Activities:**
1. Recruit an all-comers validation cohort (~15–16 EoE patients, biopsy + PBMC) without HLA-based selection; genotype each and compute their individualized epitope panel.
2. Run the Task 3 assay against each patient's own panel: expand allergen-specific CD4+ T cells; measure proliferation (SI), Th2 cytokines (IFNγ/IL-5/IL-13), activation (CD25/HLA-DR).
3. Synthesize research-grade patient-matched pMHC-II multimers loaded with that patient's top predicted cores; test target engagement (tetramer staining) and anergy induction ex vivo.
4. TCR-sequence dominant allergen-specific clones (repertoire breadth per epitope).

**Deliverable figures/tables:** response table (SI + cytokines) plotted against each patient's individualized predicted burden; tetramer-staining frequency plots; anergy-induction dose-response; TCR clonotype summary.

**GATE A (go/no-go):**
- Allergen-specific response correlates with individualized predicted epitope burden across the cohort (positive slope, target detectable response in the majority of patients).
- Patient-matched pMHC-II multimer stains ≥0.1% of CD4+ and reduces proliferation ≥3-fold on rechallenge.
- **No-go trigger:** no burden–response relationship, or no measurable multimer engagement → revisit epitope selection / prediction method before investing in reagent scale-up.

---

## PHASE 2 — Reagent Development & In Vivo Proof-of-Concept (Months 6–18)

**Goal:** Establish a scalable pMHC-II multimer reagent and demonstrate efficacy + antigen-specificity in an animal model.

**Activities:**
1. Reagent engineering: optimize MHC-II refolding, biotinylation, multimerization (tetramer → higher-valency dextramer); define potency assay (ex vivo anergy induction).
2. In vivo model: HLA-DR-transgenic or humanized mouse; experimental EoE via allergen (or OVA-epitope substitution), leveraging established models (Mishra 2001; Saito Akei 2005).
3. Efficacy study: vehicle vs pMHC-II multimer vs irrelevant-peptide multimer; endpoints = esophageal/GI eosinophil count, allergen-specific Th2 cytokines, histology.
4. Specificity check: confirm unrelated (e.g., viral-recall) T-cell responses are preserved.

**Deliverable figures/tables:** eosinophil-count reduction by arm (with n and stats); Th2-cytokine suppression; specificity panel (target vs off-target responses); reagent potency/stability sheet.

**GATE B:**
- ≥50% reduction in tissue eosinophils vs vehicle (p<0.05), specific to the correct-peptide multimer.
- Preserved unrelated immunity.
- **No-go trigger:** no eosinophil reduction, or loss of unrelated immunity (specificity failure) → return to Mechanism decision tree (add costimulation-blockade adjunct or escalate to particle/CAR approaches).

---

## PHASE 3 — IND-Enabling Studies (Months 15–30)

**Goal:** Generate the CMC, GLP toxicology, and PK/PD package required for an IND.

**Activities:**
1. **CMC:** transfer reagent to GMP; define drug substance/product specs, reference standard, stability; validate potency and purity (endotoxin, mycoplasma, adventitious agents).
2. **GLP toxicology:** repeat-dose study in a relevant species (HLA-transgenic where needed); 3–5 dose levels; TK sampling.
3. **Immunotoxicology:** confirm epitope-selective action spares protective immunity; assess cytokine-release and anti-drug-antibody potential.
4. **PK/PD:** serum half-life, tissue distribution; anergy-marker kinetics (Task 3 panel) as the PD bridge.
5. Assemble pre-IND briefing package; request FDA pre-IND meeting.

**Deliverable figures/tables:** dose–toxicity summary; TK curves; ADA incidence; PD-marker time-course; CMC specification table.

**GATE C:**
- No adverse findings at projected therapeutic multiples (acceptable safety margin).
- Reproducible GMP lots meeting specs.
- Clear PK/PD relationship.
- **No-go trigger:** narrow safety margin, ADA neutralizing efficacy, or CMC non-reproducibility → reformulate or re-engineer valency/dose.

---

## PHASE 4 — First-in-Human Phase 1b/2a (Months 30–42)

**Goal:** Establish safety and an early efficacy/PD signal in an all-comers EoE cohort, with each patient treated using their own personalized pMHC reagents.

**Design (from Task 2 & mechanism review):** randomized, placebo-controlled, **HLA-agnostic (all-comers)**; ~60 patients powered to detect a burden–response correlation of r≈0.35 (Fisher-z, α=0.05, power 0.80) with margin for covariate adjustment. No patient is excluded on HLA grounds; enrollment targets a cohort reflecting the ancestral diversity of the EoE population.

**Endpoints:**
- **Primary:** association between each patient's individualized predicted epitope burden and treatment response (regression across the full cohort, slope > 0, p<0.05), with clinically meaningful peak-eosinophil reduction (≥50%) in a substantial fraction of patients.
- **Secondary:** allergen-specific CD4+ proliferation/cytokine suppression (Task 3 ex vivo, against each patient's own panel); esophageal IL-5/IL-13 mRNA; dysphagia score; EREFS.
- **PD/mechanism:** anergy markers (CD25 down, PD-1 up); response magnitude vs individualized predicted burden.
- **Safety:** AEs (CTCAE v5), infusion reactions, ADA.
- **Equity:** response is not confined to any single ancestry or allele group; a valid personalized reagent is generated for ≥95% of enrolled patients.

**Deliverable figures/tables:** PIE waterfall (whole cohort); burden–response regression; T-cell PD time-course; enrollment-demographics/equity summary; safety summary.

**Success:** primary met with acceptable safety → advance to larger Phase 2b.

---

## Cross-Phase Risk Register

| Risk | Phase | Mitigation |
|---|---|---|
| Epitope predictions don't reflect true immunodominance | 1 | Ex vivo validation (Gate A) before reagent investment; TCR-seq confirms real clones |
| Anergy not durable (memory persists) | 1–2 | Test reversibility early; pre-plan CTLA-4 adjunct (Mechanism 3) |
| Predictor may not cover a patient's rare allele | all | Predictor spans 136 class-II alleles (DRB1/DQ/DP); 13 common DRB1 characterized against the allergen panel and cross-validated (ρ=0.79). For an uncovered allele, compute on netMHCIIpan (thousands of alleles) — the platform is per-patient, so no patient is excluded |
| Predicted epitope burden may not reflect true immunodominance | 1 | Burden is a covariate, validated against ex vivo response at Gate A before it informs anything downstream |
| In vivo model imperfectly mimics human EoE | 2 | Use established allergen/OVA EoE models; human ex vivo remains primary evidence |
| Specificity failure (protective immunity lost) | 2–3 | Explicit off-target immunity panel at Gate B and in immunotox |
| Reagent CMC not scalable | 3 | Early GMP transfer; potency assay locked in Phase 2 |

---

## Dependencies Summary

- **Task 1** (epitope panel) → peptides for reagents and assays.
- **Task 2** (HLA stratification) → patient selection in Phases 1 and 4.
- **Task 3** (T-cell assay) → Gate A readout and the clinical PD biomarker.
- **Mechanism review** → mechanism choice and the escalation ladder behind each no-go gate.

---

**Note on estimates:** phase durations overlap by design (reagent development begins during ex vivo validation). All success thresholds are pre-registered planning targets, to be finalized with statistical and regulatory review before each phase.