In a previous study, Hughes et al. [Proc. Natl. Acad. Sci. USA 93 (1996) 2065-2070] demonstrated that the amyloid peptide is able to interact with itself in a two-hybrid system and that interaction is specific. They further supported that the method could be used to define the sequences that might be important in nucleation-dependent aggregation. The sequence of the amyloid peptide can be split into four clusters, two hydrophilic (1-16 and 22-28) and two hydrophobic (17-21 and 29-42). We designed by molecular modeling and tested by the two-hybrid approach, series of mutations spread all over the sequence and changing the distribution of hydrophobicity and/or the spatial hindrance. In the two-hybrid assay, interaction of native Abeta is reproduced. Screening of mutations demonstrates that the C-domain (residues 29-40 (42)), the median domain (residues 17-22) and the N-domain (1-16) are all crucial for interaction. This demonstrates that almost all fragments of the amyloid peptide but a loop (residues 23-28) and the C-term amino acid are important for the native interaction. We support that the folded three-dimensional (3D) structure is the Abeta-Abeta interacting species, that the whole sequence is involved in that 3D fold which has a low secondary structure propensity and a high susceptibility to mutations and thus should have a low stability. The native fold of Abeta could be stabilized in Abeta-Abeta complexes which could in other circumstances facilitate the nucleation event of aggregation that leads to the formation of stable senile plaques.