Selective destruction of malignant tumor cells without damaging normal cells is an important goal for cancer chemotherapy in the 21st century. Differentiating agents that transform cancer cells to either a nonproliferating or normal phenotype could potentially be tissue-specific and avoid side effects of current drugs. However, most compounds that are presently known to differentiate cancer cells are histone deacetylase inhibitors that are of low potency or suffer from low bioavailability, rapid metabolism, reversible differentiation, and nonselectivity for cancer cells over normal cells. Here we describe 36 nonpeptidic compounds derived from a simple cysteine scaffold, fused at the C-terminus to benzylamine, at the N-terminus to a small library of carboxylic acids, and at the S-terminus to 4-butanoyl hydroxamate. Six compounds were cytotoxic at nanomolar concentrations against a particularly aggressive human melanoma cell line (MM96L), four compounds showed selectivities of > or =5:1 for human melanoma over normal human cells (NFF), and four of the most potent compounds were further tested and found to be cytotoxic for six other human cancer cell lines (melanomas SK-MEL-28, DO4; prostate DU145; breast MCF-7; ovarian JAM, CI80-13S). The most active compounds typically caused hyperacetylation of histones, induced p21 expression, and reverted phenotype of surviving tumor cells to a normal morphology. Only one compound was given orally at 5 mg/kg to healthy rats to look for bioavailability, and it showed reasonably high levels in plasma (C(max) 6 microg/mL, T(max) 15 min) for at least 4 h. Results are sufficiently promising to support further work on refining this and related classes of compounds to an orally active, more tumor-selective, antitumor drug.