**Background**
The retinoid pathway, mediated by all-trans retinoic acid (ATRA), is critical for embryonic development, and its disruption can lead to developmental toxicity. Regulatory interest in the retinoid system is growing, but no validated OECD assays currently probe this pathway. To support the development of new approach methodologies (NAMs) for developmental and reproductive toxicity (DART) testing, a comprehensive set of reference chemicals with known activity on retinoid pathway targets is needed. This study aimed to compile such a set from public databases and literature, and to assess literature support for adverse outcome pathways (AOPs) linking chemical activity to developmental defects.
**Methods**
The authors surveyed ten protein targets in the retinoid pathway: retinol binding protein (RBP), STRA6, cellular retinol binding proteins (CRBP1-3), cellular retinoic acid binding proteins (CRABP1-2), CYP26A1/B1, retinol dehydrogenase (RDH), retinal dehydrogenase (ALDH/RALDH), and retinoic acid receptors alpha, beta, and gamma (RARa, RARb, RARg). Data were extracted from Protein Data Bank (PDB), ChEMBL, ToxCast/Tox21, and PubMed via MeSH term queries. For ToxCast/Tox21, hits were limited to AC50 < 2.0 μM and curated to remove equivocal hits (fewer than three caution flags, below cytotoxicity cutoff). An updated Abstract Sifter tool was populated with the candidate chemicals and complex search terms for downstream biological queries. A case study was performed on citral, a known RALDH inhibitor, to evaluate literature support for a provisional AOP linking RALDH inhibition to limb defects (MIE: RALDH inhibition; KE: decreased ATRA synthesis, disrupted FGF8 gradient, altered DHAND expression, disrupted Hox gene expression; AO: limb defects).
**Key Results**
Approximately 280 candidate reference chemicals were identified across the ten targets. Data volume varied widely: RBP had many PDB and ChEMBL entries, while STRA6 and CYP26 had few. For RBP, key chemicals included fenretinide, A1120, and BPN-14136. For CRABP, active compounds included ATRA, TTNPB, AM80, and CD367. CYP26 inhibitors included liarozole, talarozole, ketoconazole, and several azoles. RDH inhibitors included 4-methylpyrazole, cimetidine, and ethanol. ALDH/RALDH inhibitors included disulfiram, citral, DEAB, WIN18,446, and several thiocarbamate pesticides. RARa agonists/antagonists included AM580, AM80, BMS493, and AGN 193109. RARb ligands included BPDE, BMS453, and CD2019. RARg ligands included BMS961, CD1530, and trifarotene. ToxCast/Tox21 assays covered only RARa, RARb, RARg, and the retinoid pathway (DR5 and RARE reporter assays). In the literature exploration, approximately 64% of the 280 chemicals had some connection to developmental toxicity citations, and 58 had literature linking them to limb defects. The citral case study found evidence for RALDH inhibition (MIE), decreased ATRA synthesis, and disrupted FGF signaling in facial development, but no direct citations linking citral to DHAND or ZPA activity. Citral inhibited Hoxa1 expression by 75% in two studies. In vivo, citral caused limb malformations in chick embryos and minor skeletal abnormalities in rats at oral doses >60 mg/kg.
**Clinical Implications**
This compendium of reference chemicals and the updated Abstract Sifter tool provide a foundational resource for developing and validating in vitro assays for retinoid pathway disruption, supporting the transition to animal-free DART testing. The identified chemicals can be used to calibrate assay thresholds and establish confidence in NAMs. However, the case study highlights significant data gaps: many chemicals lack evidence linking molecular initiating events to downstream key events and adverse outcomes, particularly at the cellular and tissue level. This underscores the need for complex in vitro systems (e.g., organ-on-a-chip) that can measure AOP-relevant endpoints. The patchy data coverage across targets (e.g., no ToxCast assays for RBP, STRA6, CRBP, CRABP, CYP26, RDH, or ALDH) limits the ability to assess environmental chemicals. Future work should prioritize filling these gaps and integrating metabolic competence into assays to improve the predictive value of in vitro retinoid toxicity testing.