We investigated the clinical, histologic, and molecular responses of normal human skin to all-trans-retinol (ROL) application, compared to those induced by topical all-trans-retinoic acid (RA), and measured ROL-derived metabolites. Up to 1.6% ROL, 0.025% RA in vehicle (70% ethanol/30% propylene glycol), or vehicle alone were applied in a double-blind fashion to normal buttock skin and occluded for 4 d. ROL produced from none to only trace erythema, which was clinically and statistically insignificant, whereas RA induced a significant 3.7-fold increase in erythema score compared to vehicle (n = 10, p < 0.01). However, ROL induced significant epidermal thickening (1.5-fold at 1.6% ROL, p < 0.01), similar to RA (1.6-fold at 0.025% RA, p < 0.01), relative to the vehicle. ROL, compared with vehicle, also increased mRNA levels of cellular retinoic acid binding protein (CRABP-II) and cellular retinol binding protein (CRBP) genes as determined by Northern analysis (5-6-fold and 6-7-fold, respectively) and riboprobe in situ hybridization. CRABP-II and CRBP protein levels were also higher following ROL than vehicle treatment, as measured by ligand binding (3.2-fold, p < 0.001; n = 7) and Western analysis (3.6-fold, p < 0.003; n = 6), respectively. Epidermal retinyl ester (RE) content, measured after removal of stratum corneum, rose 240-fold (p < 0.005, n = 5) by 24 h of ROL occlusion. RA content, however, was undetectable or detectable only at trace amounts in all samples obtained at 0, 6, 24, and 96 h after ROL occlusion. Detectability of RA was not correlated with ROL treatment (compared to untreated normal skin, p = 0.86) or baseline skin ROL levels (average r = -0.1, p > 0.3). These data demonstrate that ROL application 1) produces trace erythema not significantly different from vehicle, whereas RA causes erythema; 2) induces epidermal thickening and enhances expression of CRABP-II and CRBP mRNAs and proteins as does RA; 3) causes marked accumulation of retinyl ester; and 4) does not significantly increase RA levels. Taken together, the data are compatible with the idea that ROL may be a prohormone of RA, because it produces changes in skin similar to those produced by RA but without measurable RA or irritation.