Brucella melitensis Rev 1 is the best vaccine available for the prophylaxis of small ruminant brucellosis and, indirectly, for reducing human brucellosis. However, Rev 1 shows anomalously high rates of spontaneous dissociation from smooth (S) to rough (R) bacteria, the latter being inefficacious as vaccines. This S-R instability results from the loss of the O-polysaccharide. To overcome this problem, we investigated whether some recently described mechanisms promoting mutations in O-polysaccharide genes were involved in Rev 1 S-R dissociation. We found that a proportion of Rev 1 R mutants result from genome rearrangements affecting the wbo O-polysaccharide loci of genomic island GI-2 and the wbkA O-polysaccharide glycosyltransferase gene of the wbk region. Accordingly, we mutated the GI-2 int gene and the wbk IS transposase involved in those arrangements, and found that these Rev 1 mutants maintained the S phenotype and showed lower dissociation levels. Combining these two mutations resulted in a strain (Rev 2) displaying a 95% decrease in dissociation with respect to parental Rev 1 under conditions promoting dissociation. Rev 2 did not differ from Rev 1 in the characteristics used in Rev 1 typing (growth rate, colonial size, reactivity with O-polysaccharide antibodies, phage, dye and antibiotic susceptibility). Moreover, Rev 2 and Rev 1 showed similar attenuation and afforded similar protection in the mouse model of brucellosis vaccines. We conclude that mutations targeting genes and DNA sequences involved in spontaneous O-polysaccharide loss enhance the stability of a critical vaccine phenotype and complement the empirical stabilization precautions taken during S Brucella vaccine production.