# M3 repurposing validation — ranked candidate shortlist Method per candidate: mechanism fit (endometriosis biology) · evidence (ChEMBL + PubMed/clinical) · safety · novelty · confidence tag. Sources cited inline; research hypothesis-generation, never clinical advice. Updated as candidates are processed. Status: 9/9 depth-checked (2026-09-02). Input: `docs/data/repurposing_candidates.json` · digest `2026-09-02`. Primary sources for all candidates verified against PubMed/Europe PMC/ ChEMBL/ClinicalTrials.gov (2026-09-02). GTEx/HPA expression cross-reference below. --- ## Ranked shortlist (top line) | Rank | Drug → target | Verdict | Confidence | Why (one line) | |---|---|---|---|---| | 1 | **Sirolimus → FKBP4** | TOP TIER | MEDIUM-HIGH | mTOR inhibition has independent endo evidence (senescence/infertility), antiproliferative direction, chronic-use safety data | | 2 | **Cetrorelix → MRGPRX2** | VALIDATED AXIS | MEDIUM | MRGPRX2 = mast-cell pain mechanism in endo; drug already used in endo — target-validating, not novel repurposing | | 3 | Tacrolimus/Cyclosporine → FKBP4 | WATCHLIST | LOW-MEDIUM | rat-model efficacy, but immunosuppressant baggage; FKBP4 hit ≠ clinical pharmacology (FKBP12-mediated) | | 4 | **Sulfasalazine → SLC7A11** | WATCHLIST | LOW-MEDIUM | target axis real + direction now resolved (lesion-cell induction therapeutic; systemic/immune + xCT-dose caveats) | | 5 | Crizotinib/Dabrafenib → ACVR1B | WATCHLIST | LOW | ACVR1B/ALK4→aromatase mechanism is real, but cancer-TKI fertility baggage + off-target heavy | | — | Estradiol → GPER1 | ✗ WRONG DIRECTION | — | agonist of a proliferation-driving receptor — target-validating only | | — | Dinoprostone → SLCO2A1 | ✗ WRONG DIRECTION | — | it IS PGE2; therapy would need transporter *blockade* | --- ## Candidate 1 — Sulfasalazine → SLC7A11 (xCT) **Verdict: WATCHLIST — LOW-MEDIUM.** Specific xCT inhibitor → GSH depletion → ferroptosis (strong cancer preclinical, PMC7400102; Oncogene 2015, 10.1038/onc.2015.60). ⚠️ Direction of effect — **RESOLVED 2026-09-02** (`ferroptosis-direction.md`, MEDIUM confidence): ferroptosis induction in lesion cells is therapeutic (lesions evade it via MGST3/HSD11B1/FZD7–SLC7A11; forcing it shrinks lesions in 2026 models); SEMA3C's ferroptotic signaling is sub-lethal/exploitative, not evidence against induction. Remaining blockers are systemic/selectivity: xCT blockade also ferroptoses lesional CD8⁺ T cells (pro-lesion, PMID 41146213) and eutopic endometrium (fertility) — argues lesion-targeted use, and xCT-active doses carry real toxicity. Safety split: fine at RA doses; xCT-active doses (4.5–6 g) — 2010 glioma trial terminated (ISRCTN45828668, mean 7.2 AEs/patient) vs 2026 phase 1 (NCT04205357, Redox Biol 10.1016/j.redox.2026.104241) well tolerated at 3-day dosing with confirmed GSH reduction. No endo trials; novelty high. ## Candidate 2+3 — Crizotinib / Dabrafenib → ACVR1B **Verdict: WATCHLIST — LOW.** Mechanism is real and endo-specific: ACVR1B/ALK4 mediates **activin A → aromatase** (local estrogen synthesis) in endometriosis (ALK4-Smad-aromatase axis), and ALK4 haplodeficiency mitigates disease in models. But both drugs are targeted cancer TKIs: fertility/reproductive toxicity (crizotinib: ovarian follicle necrosis in rats, reversible hypogonadism), chronic-dosing burden, and their primary pharmacology (ALK/ROS1/MET; BRAF) is irrelevant to endo — off-target engagement at useful ACVR1B doses is unproven. Direction (kinase inhibition → less aromatase) is correct; drug fit is poor. Better: a selective ACVR1B inhibitor program, not these TKIs. ## Candidate 4+5+6 — Cyclosporine / Sirolimus / Tacrolimus → FKBP4 **Verdict: TOP TIER (sirolimus) — WATCHLIST (CsA/tacro).** - Mechanism nuance: FKBP4/FKBP52 is the progesterone-receptor co-chaperone that ENHANCES PR transcription (FKBP5 inhibits); FKBP4 is HOXA10-regulated in endometriosis — PR chaperone biology is central to progesterone resistance. **But** the clinical pharmacology of these drugs runs through FKBP12 (calcineurin for CsA/tacro; mTOR for sirolimus) — the ChEMBL FKBP4 hit is binding promiscuity, not their patient-facing mechanism. Rank by the drug's real pharmacology (naming note: FKBP4 gene = FKBP52 protein, FKBP5 = FKBP51 — the PR-chaperone literature uses the protein names: FKBP52 deficiency confers uterine progesterone resistance in vivo, J Clin Invest 2007 PMID 17571166; Fkbp52−/− mice develop progesterone-resistant endometriosis, Am J Pathol 2008 PMID 18988805; FKBP4 mRNA is reduced in eutopic endometrium of women with endometriosis, Reproduction 2012 PMID 22279148 + JCEM 2017 miR-29c PMID 27778641): - **Sirolimus (rapamycin):** mTOR inhibition has an independent, growing endometriosis evidence base — "mTOR inhibitors as potential therapeutics for endometriosis" narrative review (Mol Hum Reprod 2024, PMID 39579091); rapamycin shrinks lesions in mice (Exp Ther Med 2016, PMID 27347023); improves endo-associated infertility via ovarian senescence through PPARα/IGFBP2 (Mol Med Rep 2026, PMID 41170754), with a retrospective human IVF corollary (Reprod Biomed Online 2024, PMID 37914557: 168 patients, 80 rapamycin-treated); antiproliferative direction; chronic-use safety established (transplant + LAM, EMA Rapamune). No registered endometriosis trial of rapamycin/sirolimus/any mTOR inhibitor on ClinicalTrials.gov (checked 2026-09-02). **TOP TIER.** - **CsA/tacrolimus:** efficacy in rat endometriosis model (anti- inflammatory/immunomodulatory), but chronic immunosuppression (infection, nephrotoxicity) is heavy for a benign chronic disease. Watchlist only. ## Candidate 7 — Estradiol → GPER1 **Verdict: ✗ WRONG DIRECTION — target-validating only.** GPER expression is increased in endometriosis (Plante 2012, cited 141×); GPER agonist G-1 stimulates endometrial proliferation. Estradiol is the endogenous GPER **agonist** — activating a proliferation-driving receptor in a proliferative disease is anti-therapeutic. The hit **validates GPER1 as a target** (antagonist development), not estradiol as a drug. Textbook case for the direction-of-effect review step. ## Candidate 8 — Cetrorelix → MRGPRX2 **Verdict: VALIDATED AXIS (target-validating, not repurposing) — MEDIUM.** Two findings converge: (1) MRGPRX2 mediates mast-cell-induced endometriosis **pain** via HBD-2 → MRGPRX2 → histamine → HRH1/TRPV1 sensory-neuron sensitization — MRGPRX2⁺ mast-cell density is increased in lesions and mast-cell/MRGPRX2 loss or H1 blockade relieves hyperalgesia in models (FASEB J **2025**, PMID 40600649; earlier draft said 2026 — corrected); (2) cetrorelix (Cetrotide, 3 mg once weekly × 8 weeks) is an established GnRH-antagonist endometriosis treatment — regimen documented in small 2002 reports (Reprod Biomed Online, PMIDs 12537785 + 12470539); a registered phase-2 used a different SR single-dose design and posted no results (NCT00244452). The ChEMBL engagement is **measured but agonistic**: EC50 617/813 nM (pChEMBL 6.21/6.09, Ca²⁺ + β-arrestin assays; Lansu et al., Nat Chem Biol 2017, PMID 28288109) — i.e. cetrorelix *activates* MRGPRX2; no clinical link between cetrorelix and mast-cell action exists. As repurposing this is moot; as **target validation** it is strong: MRGPRX2 **antagonism** is a specific, non-hormonal pain target worth developing for endo pain. ## Candidate 9 — Dinoprostone (PGE2) → SLCO2A1 **Verdict: ✗ WRONG DIRECTION — target-validating only.** SLCO2A1 is the prostaglandin transporter; endometriotic tissue overproduces PGE2 and PGE2 drives lesion growth/VEGF. Dinoprostone IS PGE2 (agonist) — therapy would need SLCO2A1 **blockade** or PGE2 synthesis inhibition. Like estradiol: target validated (PG transport axis), drug is the wrong tool. ## Expression cross-reference (GTEx v8 / Human Protein Atlas, 2026-09-02) Bulk-tissue expression for the ranked targets (HPA tissue-RNA consensus nTPM; GTEx v8 median TPM — GTEx has no endometrium, nearest proxy = uterus): | Gene | Endometrium nTPM (HPA) | Uterus TPM (GTEx v8) | Lesional evidence | Verdict | |---|---|---|---|---| | FKBP4 | 30.6 (IHC high) | 43.5 | mRNA ↓ in eutopic EM endometrium (PMIDs 22279148, 27778641) | Plausible — present & mechanistically linked | | MRGPRX2 | 0.0 (not detected) | 0.06 | MRGPRX2⁺ mast cells ↑ in lesions (PMID 40600649) | Plausible only via lesional mast cells | | SLC7A11 | 0.2 (not detected) | 0.12 | SLC7A11 ↓ in ectopic lesions (PMID 42678895); xc⁻/erastin concept (PMID 41001371) | Weak expression support at disease site | Interpretation: FKBP4 and (via lesional mast cells) MRGPRX2 pass the "is the target there?" test. SLC7A11 is near-undetectable in normal reproductive tissue and *down*-regulated in lesions — sulfasalazine's on-target rationale at the disease site is the thinnest of the ranked set, one reason it stays WATCHLIST LOW-MEDIUM. Sources: proteinatlas.org ENSG00000004478-FKBP4/tissue · ENSG00000183695-MRGPRX2/tissue · ENSG00000151012-SLC7A11/tissue; gtexportal.org/home/gene/<ENSG>. --- ## Phase 0.5 lesion-expression corroboration (GEO, 2026-09-03/04) Lesion-level transcriptomic reanalysis (full methods: `docs/research/virtual/phase0.5.md`) adds disease-site expression evidence to the GTEx/HPA cross-reference above — the GTEx/HPA table reflects *normal* tissue; these are *lesional* measurements. | Dataset | Design | Findings relevant to M3 | |---|---|---| | GSE282532 (PR #13/#18) | endometrioma bulk RNA-seq, 5 patient-paired eutopic vs ectopic | **FKBP4 2.2× DOWN** in ectopic (FC 0.45, all 5 pairs — PR-resistance consistent); **SLC7A11 2× down**, ACSL4 2.4× down (ferroptosis-evasion via effector suppression); mast-cell markers CPA3/TPSAB1/TPSB2 **11–15× up** | | GSE247695 (PR #19) | scRNA-seq, 4 pairs | FKBP4 down in lesion **stroma** (2 datasets agree); SLC7A11 down (stroma/endothel); MRGPRX2 below scRNA detection (dropout — inconclusive) | | GSE263897 (PR #20) | GeoMx spatial, 5 pairs, 60 segments | FKBP4 stroma down (0.89, 4/5) + **epithelium up (1.31, 5/5)** — compartment split replicated; SLC7A11 no constitutive up-regulation; mast-cell markers + MRGPRX2 at negative-probe background → **enrichment is endometrioma-specific, absent in superficial peritoneal lesions** (resolves the GSE247695 discrepancy as lesion type, not method) | Reading per ranked candidate: - **Sirolimus → FKBP4 (TOP TIER):** FKBP4 is present and disease-relevant in lesions — *down*-regulation in lesion tissue/stroma is the PR-resistance signature (FKBP4↓ → weaker PR signalling), consistent with the chaperone hypothesis the ChEMBL hit pointed at. Expression data do not contradict FKBP4 engagement; Phase 1.0 docking (PR #22) shows rapamycin fits FKBP52 −6.38 vs FKBP12 −7.06. - **Sulfasalazine → SLC7A11 (WATCHLIST):** SLC7A11 is *low/suppressed* in lesion tissue (bulk + scRNA + spatial agree) — ferroptosis-evasion reads as effector suppression, which *supports* the xCT-inhibition direction at the expression level. The thin normal-tissue expression noted above is therefore not the blocker; selectivity/toxicity remain the WATCHLIST issue (unchanged). - **Cetrorelix → MRGPRX2 (VALIDATED AXIS):** the mast-cell/MRGPRX2 pain axis is confirmed in **endometrioma** (GSE282532: 11–15× mast markers) but **not in superficial peritoneal lesions** (GSE247695/GSE263897) — lesion-type caveat now explicit; IHC stratified by lesion type is the definitive check. This tempers MRGPRX2-antagonist priority for peritoneal disease, not the target itself. --- ## Next steps (DoD status) 1. ✅ Knowledge pages (in this PR): `entities/sirolimus.md` + `concepts/mrgprx2-pain.md` (SCHEMA-compliant, confidence tags) + knowledge index/log + llms.txt + wiki.html 2. ✅ GTEx/HPA expression cross-reference (section above) 3. ✅ Phase 0.5 GEO lesion-expression corroboration folded in (2026-09-04) — Phase 1.0 docking complete (PR #22, 2026-09-04): sulfasalazine→xCT −8.52, rapamycin→FKBP52 −6.38; cetrorelix→MRGPRX2 deferred to Phase 2 (tool boundary) 4. ⏳ Share ranked list with human (wckdboy) for any clinical-interest decision — research output, not medical advice (see PR #12) *Validation by Percival (Hermes Agent), 2026-09-02. All sources above are public (PubMed/PMC/Europe PMC/ClinicalTrials.gov).*