Mycobacterium smegmatis Dps degrades spontaneously into a species in which 16 C-terminal residues are cleaved away. A second species, in which all 26 residues constituting the tail were deleted, was cloned, expressed and purified. The first did not bind DNA but formed dodecamers like the native protein, while the second did not bind to DNA and failed to assemble into dodecamers, indicating a role in assembly also for the tail. In the crystal structure of the species without the entire C-terminal tail the molecule has an unusual open decameric structure resulting from the removal of two adjacent subunits from the original dodecameric structure of the native form. A Dps dodecamer could assemble with a dimer or one of two trimers (trimer-A and trimer-B) as intermediate. Trimer-A is the intermediate species in the M. smegmatis protein. Estimation of the surface area buried on trimerization indicates that association within trimer-B is weak. It weakens further when the C-terminal tail is removed, leading to the disruption of the dodecameric structure. Thus, the C-terminal tail has a dual role, one in DNA binding and the other in the assembly of the dodecamer. M. smegmatis Dps also has a short N-terminal tail. A species with nine N-terminal residues deleted formed trimers but not dodecamers in solution, unlike wild-type M. smegmatis Dps, under the same conditions. Unlike in solution, the N-terminal mutant forms dodecamers in the crystal. In native Dps, the N-terminal stretch of one subunit and the C-terminal stretch of a neighboring subunit lock each other into ordered positions. The deletion of one stretch results in the disorder of the other. This disorder appears to result in the formation of a trimeric species of the N-terminal deletion mutant contrary to the indication provided by the native structure. The ferroxidation site is intact in the mutants.