The Abeta1-42 peptide that is overproduced in Alzheimer's disease (AD) from a large precursor protein has a normal amino acid sequence but, when liberated, misfolds at neutral pH to form "protofibrils" and fibrils that are rich in beta-sheets. We find that these protofibrils or fibrils are toxic to certain neuronal cells that carry Ca-permeant alpha-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid (AMPA) receptors. Disrupting the structure of the Abeta1-42 fibrils and protofibrils might lead to the discovery of molecules that would be very useful in the treatment of AD. A high-throughput screen of a library of >3,000 small molecules with known "biological activity" was set up to find compounds that efficiently decrease the beta-sheet content of aggregating Abeta1-42. Lead compounds were characterized by using thioflavin T (ThT) as a beta-sheet assay. The most effective of six compounds found was 4,5-dianilinophthalimide (DAPH) under the following conditions: DAPH at low micromolar concentrations abolishes or greatly reduces previously existing fully formed Abeta1-42 fibrils, producing instead amorphous materials without fibrils but apparently containing some protofibrils and smaller forms. Coincubation of the Abeta1-42 peptide with DAPH produces either amorphous materials or empty fields. Coincubation of DAPH and Abeta1-42 greatly reduces the beta-sheet content, as measured with ThT fluorescence, and produces a novel fluorescent complex with ThT. When the Abeta1-42 peptide was coincubated with DAPH at very low micromolar concentrations, the neuronal toxicity mentioned above (Ca(2+) influx) was eliminated. Clearly, DAPH is a promising candidate for AD therapy.