The beta-amyloid peptide (Abeta) is a normal product of the proteolytic processing of its precursor (beta-APP). Normally, it elicits a very low humoral immune response; however, the aggregation of monomeric Abeta to form fibrillar Abeta amyloid creates a neo-epitope, to which antibodies are generated. Rabbits were injected with fibrillar human Abeta(1-42), and the resultant antibodies were purified and their binding properties characterized. The antibodies bound to an epitope in the first eight residues of Abeta and required a free amino terminus. Additional residues did not affect the affinity of the epitope as long as the peptide was unaggregated; the antibody bound Abeta residues 1-8, 1-11, 1-16, 1-28, 1-40, and 1-42 with similar affinities. In contrast, the antibodies bound approximately 1000-fold more tightly to fibrillar Abeta(1-42). Their enhanced affinity did not result from their bivalent nature: monovalent Fab fragments exhibited a similar affinity for the fibrils. Nor did it result from the particulate nature of the epitope: monomeric Abeta(1-16) immobilized on agarose and soluble Abeta(1-16) exhibited similar affinities for the antifibrillar antibodies. In addition, antibodies raised to four nonfibrillar peptides corresponding to internal Abeta sequences did not exhibit enhanced affinity for fibrillar Abeta(1-42). Antibodies directed to the C-terminus of Abeta bound poorly to fibrillar Abeta(1-42), which is consistent with models where the carboxyl terminus is buried in the interior of the fibril and the amino terminus is on the surface. When used as an immunohistochemical probe, the antifibrillar Abeta(1-42) IgG exhibited enhanced affinity for amyloid deposits in the cerebrovasculature. We hypothesize either that the antibodies recognize a specific conformation of the eight amino-terminal residues of Abeta, which is at least 1000-fold more favored in the fibril than in monomeric peptides, or that affinity maturation of the antibodies produces an additional binding site for the amino-terminal residues of an adjacent Abeta monomer. In vivo this specificity would direct the antibody primarily to fibrillar vascular amyloid deposits even in the presence of a large excess of monomeric Abeta or its precursor. This observation may explain the vascular meningeal inflammation that developed in Alzheimer's disease patients immunized with fibrillar Abeta. Passive immunization with an antibody directed to an epitope hidden in fibrillar Abeta and in the transmembrane region of APP might be a better choice in the search for an intervention to remove Abeta monomers without provoking an inflammatory response.