Amyloid-beta (Abeta) peptides are implicated in the neurodegeneration of Alzheimer's disease (AD). We previously investigated the mechanism of neurotoxicity of Abeta and found that human Abeta (huAbeta) binds and depletes heme, forming an Abeta-heme complex with peroxidase activity. Rodent Abeta (roAbeta) is identical to huAbeta, except for three amino acids within the proposed heme-binding motif (Site-H). We studied and compared heme-binding between roAbeta and huAbeta. Unlike roAbeta, huAbeta binds heme tightly (K(d)=140+/-60 nM) and forms a peroxidase. The plot of bound (huAbeta-heme) vs. unbound heme fits best to a two site binding hyperbola, suggesting huAbeta possesses two heme-binding sites. Consistently, a second high affinity heme-binding site was identified in the lipophilic region (site-L) of huAbeta (K(d)=210+/-80 nM). The plot of (roAbeta-heme) vs. unbound heme, on the other hand, was different as it fits best to a sigmoidal binding curve, indicating different binding and lower affinity of roAbeta for heme (K(d)=1 microM). The effect of heme-binding to site-H on heme-binding to site-L in roAbeta and huAbeta is discussed. While both roAbeta and huAbeta form aggregates equally, rodents lack AD-like neuropathology. High huAbeta/heme ratio increases the peroxidase activity. These findings suggest that depletion of regulatory heme and formation of Abeta-heme peroxidase contribute to huAbeta's neurotoxicity in the early stages of AD. Phylogenic variations in the amino acid sequence of Abeta explain tight heme-binding to huAbeta and likely contribute to the increased human susceptibility to AD.