Aggregation of the amyloid-beta (Abeta) peptide is considered a central event in the pathogenesis of Alzheimer's disease (AD). In order to bypass methodological bias related to a variety of impurities commonly present in typical preparations of synthetic Abeta, we developed a simple, generally applicable method for recombinant production of human Abeta and Abeta variants in Escherichia coli that provides milligram quantities of Abeta in very high purity and yield. Amyloid fibril formation in vitro by human Abeta1-42, the key amyloidogenic Abeta species in AD, was completed threefold faster with recombinant Abeta1-42 compared to synthetic preparations. In addition, recombinant Abeta1-42 was significantly more toxic to cultured rat primary cortical neurons, and it was more toxic in vivo, as shown by strongly increased induction of abnormal phosphorylation of tau and tau aggregation into neurofibrillary tangles in brains of P301L tau transgenic mice. We conclude that even small amounts of impurities in synthetic Abeta-including a significant fraction of racemized peptides that cannot be avoided due to the technical limitations of peptide synthesis--prevent or slow Abeta incorporation into the regular quaternary structure of growing beta-amyloid fibrils. The results validate the use of recombinant Abeta1-42 for both in vitro and in vivo studies addressing the mechanisms underlying Abeta aggregation and its related biological consequences for the pathophysiology, therapy, and prevention of AD.