The relationship between amyloid and toxic species is a central problem since the discovery of amyloid structures in different diseases. Despite intensive efforts in the field, the deleterious species remains unknown at the molecular level. This may reflect the lack of any structure-toxicity study based on a genetic approach. Here we show that a structure-toxicity study without any biochemical prerequisite can be successfully achieved in yeast. A PCR mutagenesis of the amyloid domain of HET-s leads to the identification of a mutant that might impair cellular viability. Cellular and biochemical analyses demonstrate that this toxic mutant forms GFP-amyloid aggregates that differ from the wild-type aggregates in their shape, size and molecular organization. The chaperone Hsp104 that helps to disassemble protein aggregates is strictly required for the cellular toxicity. Our structure-toxicity study suggests that the smallest aggregates are the most toxic, and opens a new way to analyze the relationship between structure and toxicity of amyloid species.