# Celiac Disease Omics: Immune Targets to Preserve the Villi ### A tolerability-nuanced target-discovery report for a diet-adherent, biologic-skeptical patient community **Scope.** This report mines public celiac-disease omics to identify immune mechanisms whose modulation could protect the intestinal villi from the damage caused by inadvertent gluten exposure — and prioritizes them through a *patient-preference lens*. Unlike eosinophilic esophagitis, where a systemic biologic (dupilumab) achieved rapid uptake, the celiac community is largely managed by a gluten-free diet (GFD) and, by patient-reported evidence, is more cautious about drugs with systemic off-target effects. The prioritization therefore favors **oral, gut-restricted, low-side-effect, repurposable** agents (Tier 1) while separately surfacing **novel high-value biology** that may require new modalities and serves the refractory population (Tier 2). --- ## 1. The patient-preference rationale (why tolerability leads) A 372-patient survey (Branchi et al., *Digestion* 2016;93(2):160–166) found the GFD is accepted by **88%** of patients, yet **65%** would still want a complementary therapy — and among those, the preferred modalities were **enzymes (145 subjects) and vaccines (111 subjects)**, i.e. low-burden, non-immunosuppressive options; patients requesting a drug had *lower* quality-of-life and health-status scores (p=0.003, p=0.005). The FDA's own guidance ("Celiac Disease: Developing Drugs for Adjunctive Treatment to a Gluten-Free Diet") frames drugs as **adjuncts to the GFD**, with histologic endpoints (villous-height:crypt-depth ratio, intraepithelial lymphocytosis; modified Marsh-Oberhuber) and a residual-damage target population defined by persistent histologic inflammation despite diet. **Implication for target selection.** The ideal celiac villi-protection drug is taken orally, acts locally in the gut mucosa, spares systemic immunity, and — for fastest translation — repurposes an agent with an established safety record. That is the explicit weighting of the scoring rubric below. --- ## 2. What the data say: the villous-destruction axis is an IFN-γ / JAK-STAT / IL-15-NKG2D program **Datasets.** From 32 human celiac GEO datasets (17 relevant after curation), three whole-mucosa case-control studies formed the meta-analysis pool: **GSE131705 (33 celiac / 21 control), GSE164883 (26 / 22, Marsh-graded), GSE146190 (6 / 5)**. Single-cell validation used **GSE277276** (author-annotated; a balanced 10-subject, 91,252-cell subset). **Cross-study meta-signature (random-effects, DerSimonian-Laird; 17,418 genes; 107 high-confidence).** The active-CeD vs control mucosal signature is unambiguously an interferon-driven cytotoxic program: | Gene | Pooled log2FC | padj | Role in villous destruction | |------|--------------:|------|------------------------------| | STAT1 | +1.89 | 3.3e-29 | IFN signal integrator; top TF regulator | | CXCL11 | +2.52 | 9e-3 | IFN-induced T-cell chemokine | | TGM2 | +1.37 | 1.3e-2 | Transglutaminase-2 — deamidates gluten, autoantigen | | IFNG | +1.19 | 1.3e-2 | Dominant effector cytokine | | MICB | +0.70 | 5.3e-11 | Enterocyte stress ligand (NKG2D target) | | IL15 | +0.30 | ns (bulk) | IEL survival/activation cytokine | **Pathway/regulator enrichment.** The up-regulated program is dominated by **Interferon-γ response (Hallmark q=9.7e-40, the single strongest term)** and interferon/cytokine signaling, plus a G2-M/E2F cell-cycle module (crypt hyperplasia). The down-regulated program is loss of enterocyte metabolic function (xenobiotic/fatty-acid/lipid metabolism) — the molecular signature of villous atrophy. Upstream transcriptional regulators: **IRF1 (q=5.9e-12), STAT1 (q=3.2e-10), RELA/NFKB1** — nominating the **IFN-γ→JAK-STAT axis as the master, and orally druggable, control point.** **Single-cell compartment resolution (GSE277276).** Each node localizes to exactly the cell type its biology predicts: - **IL-15** — stromal/endothelial (trans-presentation to IELs), not epithelial-autonomous. - **NKG2D (KLRK1)** — NK CD16− and γδ T-IEL (TRDC+): the killer receptor. - **IFN-γ, granzyme, perforin** — NK CD16+ and CD8+/γδ IELs: the cytotoxic effectors. - **ITGAE (CD103), CD160** — mark tissue-resident IEL populations. - **IFN-γ is up-regulated in the cytotoxic IEL/NK compartment in active disease** (pseudobulk +0.35, p=0.095 at n=5 vs 5 subjects — directionally strong, sample-size-limited). - Epithelium shows the damage readout: APOA4/SI/differentiation genes down; enterocyte fraction 30%→15%. *Caveat (flagged throughout):* in droplet scRNA-seq the massive plasma-cell infiltrate dilutes the IEL fraction and IELs capture poorly, so IEL *fraction* appears flat/lower even though absolute IEL density rises by histology. The reproducible per-cell signal is the **IFN-γ effector program**, not cell-count changes. Granzyme/perforin per-cell means are baseline-high and dilute in the expanded active pool. --- ## 3. TIER 1 — Repurposing-first: oral, low-side-effect villi protection (diet-adherent population) These targets are drugged by **approved or late-clinical oral small molecules** and map onto the meta-signature's master axis. They are the fastest, most tolerability-aligned path to protecting villi against accidental gluten exposure. **1. TG2 (TGM2) — oral transglutaminase-2 inhibitor.** *The lead.* TG2 both generates the deamidated gluten epitopes that drive the DQ2.5/DQ8 T-cell response (cross-validated by our pMHC stream) and is up in all three studies (+1.37). **ZED1227 is an oral TG2 inhibitor in celiac Phase 2** (NCT07298343 recruiting; earlier Phase 1 NCT02679014, verified via ClinicalTrials.gov) — gut-luminal action, no systemic immunosuppression. This class of drug has been reported to reduce gluten-induced mucosal damage in trials; the specific efficacy magnitude should be confirmed against the primary trial publications. Highest tolerability-strategy fit. **2. JAK1 — oral, ideally gut-selective JAK inhibitor.** JAK1 transduces both IFN-γ and IL-15Rγ signaling — the two central villous-destruction cytokines converge here, and STAT1 (JAK1's substrate) is the highest-confidence meta node. **14 approved oral JAK inhibitors** exist (upadacitinib, filgotinib, abrocitinib…); critically, **gut-selective JAK inhibitors (izencitinib, TD-1473) were designed for minimal systemic exposure** — the tolerability-optimal form. *Class caveat:* systemic pan-JAK inhibitors carry a boxed warning (infection, thrombosis); gut-restriction is the mitigation. **3. TYK2 — oral allosteric TYK2 inhibitor (druggability/tolerability adjunct, not DE-supported).** *Disclosure:* unlike JAK1/JAK2, TYK2 does **not** meet the meta-analysis bar (pooled log2FC +0.16, padj 0.64, discordant across the three studies) — it is listed for its mechanistic role (completing the IFN-α/β arm that feeds STAT1) and its best-in-class oral safety, not for differential-expression evidence. **Deucravacitinib is an allosteric oral TYK2 inhibitor, approved (psoriasis), with no JAK boxed warning.** Consider it only as a JAK-sparing alternative if the class safety of JAK1/2 inhibition is a concern. *Demoted from Tier 1 despite oral-drug hits:* PSMB9 (bortezomib/ixazomib — systemic proteasome-inhibitor chemotherapy, unacceptable side-effect profile) and NOS2 (tilarginine — clinically failed). --- ## 4. TIER 2 — Novel biology: the IL-15/NKG2D killing machine (incl. refractory population) These are the mechanistically central nodes of villous destruction that **lack an approved oral drug** and require new or repurposed biologic/novel modalities. They carry the highest biological value — and are the leading strategy for **refractory celiac disease type II (RCDII)**, where aberrant IEL clones drive damage independent of diet. **1. IL-15 — anti-IL-15 / anti-IL-15Rβ (gut-restricted biologic or oral IL-15Rβ SM).** The apex cytokine: drives IEL survival, NKG2D induction, and cytotoxicity; single-cell shows stromal trans-presentation to IELs. **AMG-714/Ordesekimab (anti-IL-15) reached Phase 2b in celiac; CALY-002 Phase 1b.** IL-15 is *the* driver of RCDII — the leading refractory strategy. A gut-restricted format would resolve the systemic-tolerability concern. **2. NKG2D (KLRK1) — anti-NKG2D or NKG2D-MIC blocker.** The activating receptor on cytotoxic IELs/NK that engages stress ligands on enterocytes to trigger epithelial killing. Blocking it protects the villi at the effector step. No celiac-specific drug yet — novel opportunity; also RCDII-relevant. **3. MIC-A/B — anti-MICA/B biologic.** The enterocyte "kill-me" stress ligand (MICB +0.70, padj ~0), induced by IL-15/IFN-γ. Anti-MICA/B antibodies exist in oncology (repurposable biology); blocking MIC shedding/engagement protects epithelium without systemic immunosuppression. **4. IFN-γ — anti-IFN-γ (gut-restricted).** The dominant effector cytokine and strongest enrichment signal. Emapalumab (anti-IFN-γ, approved for HLH) is repurposable but systemic; a gut-restricted format, or upstream JAK blockade (Tier 1), is preferred. **5. STAT1 — validates the axis; direct drugging is novel (degrader/PROTAC).** The highest-confidence meta node and master TF; intracellular and hard to drug directly, which is precisely why the **JAK-STAT upstream handle (Tier 1) is the pragmatic route.** --- ## 5. Strategic synthesis - **The villous lesion is one coherent circuit:** gluten → TG2-deamidation → DQ-restricted CD4 help → **IL-15 + IFN-γ → JAK-STAT → cytotoxic IEL/NK (NKG2D-MIC) → enterocyte death → villous atrophy.** Every stage is a target; the data rank them. - **For the diet-adherent majority who want low-burden protection against accidental exposure:** an **oral TG2 inhibitor (ZED1227-class)** or a **gut-selective JAK/TYK2 inhibitor** is the tolerability-aligned lead — local action, oral dosing, repurposable safety record. - **For refractory patients and maximal biological leverage:** the **IL-15/NKG2D-MIC killing axis** is the novel, high-value frontier, with anti-IL-15 already clinically de-risked in celiac. - **Cross-stream convergence:** TG2 links this omics analysis to our pMHC/antigen stream (it makes the epitopes *and* scores as a Tier-1 drug target), and the IL-15→NKG2D program is the mechanistic bridge from HLA-restricted presentation to actual tissue destruction. --- ## 6. Deliverables | Artifact | Contents | |----------|----------| | celiac_dataset_inventory.csv + landscape.png | 32 datasets classified; 17 relevant | | celiac_meta_signature.csv | 17,418-gene cross-study active-CeD vs control signature | | celiac_meta_analysis.png | Meta volcano + per-study concordance + villous-axis forest | | celiac_sc_validation.png (+ 3 CSVs, slim h5ad) | Single-cell compartment resolution of the axis (GSE277276) | | celiac_enrichment.csv + .png | IFN-γ/JAK-STAT pathway & TF-regulator enrichment | | celiac_target_druggability.csv + .png | 61 targets scored on tolerability rubric + Open Targets | | celiac_target_shortlist_tier1.csv | Repurposing-first oral villi-protection shortlist | | celiac_target_shortlist_tier2.csv | Novel-biology / refractory shortlist | | celiac_tolerability_rubric.md | Scoring axes + patient-preference basis | *Guardrails observed:* (1) HLA/genetics are necessary-not-sufficient (DQ2.5 in 25–40% of the population, ~3% develop celiac) — targets are mucosal-effector, not genetic-risk, nodes; (2) presentation ≠ pathology (the pMHC guardrail); (3) single-cell IEL fractions carry a droplet-dropout caveat — the validated signal is the per-cell IFN-γ effector program, not cell counts; (4) all meta and enrichment statistics are computed from the three-study pool with random-effects pooling and BH-FDR, not from any single study.