Aortic medial amyloid (AMA) is the most common localized human amyloid, occurring in virtually all of the Caucasian population over the age of 50. The main protein component of AMA, medin, readily assembles into amyloid-like fibrils in vitro. Despite the prevalence of AMA, little is known about the self-assembly mechanism of medin or the molecular architecture of the fibrils. The amino acid sequence of medin is strikingly similar to the sequence of the Alzheimer's disease (AD) amyloid-beta (Aβ) polypeptides around the structural turn region of Aβ where mutations associated with familial, early onset AD, have been identified. D25 and K30 of medin align with residues D23 and K28 of Aβ that are known to form a stabilizing salt bridge in some fibril morphologies. Here we show that substituting D25 of medin with asparagine (D25N) impedes assembly into fibrils and stabilizes non-cytotoxic oligomers. Wild-type medin, by contrast, aggregates into β-sheet rich amyloid-like fibrils within 50 h. A structural analysis of wild-type fibrils by solid-state NMR suggests a molecular repeat unit comprising at least two extended β-strands, separated by a turn stabilized by a D25-K30 salt-bridge. We propose that D25 drives the assembly of medin by stabilizing the fibrillar conformation of the peptide, and is thus reminiscent of the influence of D23 on the aggregation of Aβ. Pharmacological comparisons of wild-type medin and D25N will help to ascertain the pathological significance of this poorly under-stood protein.