ABSTRACT: The inhibitory mechanism of serpins (serine protease inhibitors) requires a folding pathway that results in a metastable native state, and not the most stable state as dictated by the Anfinsen principle. It is thus unsurprising that point mutations in serpins often lead to the bypassing of the metastable conformation in favor of a stable but inactive product. One interesting consequence is the intracellular accumulation of serpins by a process known as polymerization. Serpin polymerization underlies a diverse set of loss-of-function and gain-of-function disorders including neurodegeneration, thrombosis, emphysema, cirrhosis and angioedema. Previous work has established that polymerization proceeds via a folding intermediate (denoted M*) and involves the expansion of the main beta-sheet A to the stable six-stranded form. Over the last two decades the 'loop-sheet' hypothesis has gained wide acceptance. In this mechanism the reactive centre loop (RCL) of one serpin monomer inserts into the beta-sheet A of another (in trans), in a manner similar to what is seen for monomeric serpin after proteolytic cleavage of the RCL. We evaluated the loop-sheet mechanism by combined molecular modelling and biochemical approaches using the archetypal serpin alpha1-antitrypsin, and conclude that the loop-sheet model cannot account for serpin polymerization in vitro or in cells. We recently solved the crystal structure of a serpin dimer linked by an extensive domain-swap involving the incorporation of two strands of the central beta-sheet A. This structure suggested a radical new mechanism of polymerization - large-scale domain swapping. In order to determine if this or other mechanisms underlie alpha1-antitrypsin polymerization in vivo, we produced and crystallized a trimeric form of alpha1-antitrypsin that is recognized by an antibody specific for the pathological polymer. The structure of the trimer revealed a polymeric linkage mediated by yet another domain swap - the C-terminal 34 residues. Disulfide trapping and structural studies on polymers formed in yeast and other cells conclusively demonstrate that runaway C-terminal domain swapping underlies polymerization of the common Z-variant of alpha1-antitrypsin in vivo. In addition, an unpublished structure of the Z-variant shows the molecular defect that leads to polymerization, and suggests strategies to rescue folding using chemical chaperones.