A debilitating complication of long-term hemodialysis is the deposition of beta-2-microglobulin (beta2m) as amyloid plaques in the joint space. We have recently shown that Cu(2+) can be a contributing, if not causal, factor at concentrations encountered during dialysis therapy. The basis for this effect is destabilization and incorporation of beta2m into amyloid fibers upon binding of Cu(2+). In this work, we demonstrate that while beta2m binds Cu(2+) specifically in the native state, it is binding of Cu(2+) by non-native states of beta2m which is responsible for destabilization. Mutagenesis of potential coordinating groups for Cu(2+) shows that native state binding of Cu(2+) is mediated by residues and structures that are different than those which bind in non-native states. An increased affinity for copper by non-native states compared to that of the native state gives rise to overall destabilization. Using mass spectrometry, NMR, and fluorescence techniques, we show that native state binding is localized to H31 and W60 and is highly specific for Cu(2+) over Zn(2+) and Ni(2+). Binding of Cu(2+) in non-native states of beta2m is mediated by residues H13, H51, and H84, but not H31. Although denatured beta2m has characteristics of a globally unfolded state, it nevertheless demonstrates the following strong specificity of binding: Cu(2+) > Zn(2+) >> Ni(2+). This requires the existence of a well-defined structure in the unfolded state of this protein. As Cu(2+) effects are reported in many other amyloidoses, e.g., PrP, alpha-synuclein, and Abeta, our results may be extended to the emerging field of divalent ion-associated amyloidosis.