Familial amyotrophic lateral sclerosis (FALS) is linked to over 90 point mutations in superoxide dismutase-1 (SOD1), a dimeric metalloenzyme. The postmortem FALS brain is characterized by SOD1 inclusions in the motor neurons of regions in which neuronal loss is most significant. These findings, together with animal modeling studies, suggest that aggregation of mutant SOD1 produces a pathogenic species. We demonstrate here that a mutant form of SOD1 (A4V) that is linked to a particularly aggressive form of FALS aggregates in vitro, while wild-type SOD1 (WT) is stable. Some A4V aggregates resemble amyloid pores formed by other disease-associated proteins. The WT dimer is significantly more stable than the A4V dimer, suggesting that dimer dissociation may be the required first step of aggregation. To test this hypothesis, an intersubunit disulfide bond between symmetry-related residues at the A4V dimer interface was introduced. The resultant disulfide bond (V148C-V148C') eliminated the concentration-dependent loss of enzymatic activity of A4V, stabilized the A4V dimer, and completely abolished aggregation. A drug-like molecule that could stabilize the A4V dimer could slow the onset and progression of FALS.