Machado-Joseph disease (MJD; also called spinocerebellar ataxia type 3) is a dominantly inherited late-onset neurodegenerative disorder caused by expansion of polyglutamine (polyQ)-encoding CAG repeats in the MJD1 gene (also known as ATXN3). Proteolytic liberation of highly aggregation-prone polyQ fragments from the protective sequence of the MJD1 gene product ataxin 3 (ATXN3) has been proposed to trigger the formation of ATXN3-containing aggregates, the neuropathological hallmark of MJD(1-5). ATXN3 fragments are detected in brain tissue of MJD patients and transgenic mice expressing mutant human ATXN3(Q71)(6), and their amount increases with disease severity, supporting a relationship between ATXN3 processing and disease progression. The formation of early aggregation intermediates is thought to have a critical role in disease initiation(7,8), but the precise pathogenic mechanism operating in MJD has remained elusive(9). Here we show that L-glutamate-induced excitation of patient-specific induced pluripotent stem cell (iPSC)-derived neurons initiates Ca2+-dependent proteolysis of ATXN3 followed by the formation of SDS-insoluble aggregates. This phenotype could be abolished by calpain inhibition, confirming a key role of this protease in ATXN3 aggregation. Aggregate formation was further dependent on functional Na+ and K+ channels as well as ionotropic and voltage-gated Ca2+ channels, and was not observed in iPSCs, fibroblasts or glia, thereby providing an explanation for the neuron-specific phenotype of this disease. Our data illustrate that iPSCs enable the study of aberrant protein processing associated with late-onset neurodegenerative disorders in patient-specific neurons.