Classical brucellosis vaccines induce antibodies to the O-polysaccharide section of the lipopolysaccharide that interfere in serodiagnosis. Brucella rough (R) mutants lack the O-polysaccharide but their usefulness as vaccines is controversial. Here, Brucella melitensis R mutants in all main lipopolysaccharide biosynthetic pathways were evaluated in sheep in comparison with the reference B. melitensis Rev 1 vaccine. In a first experiment, these mutants were tested for ability to induce anti-O-polysaccharide antibodies, persistence and spread through target organs, and innocuousness. Using the data obtained and those of genetic studies, three candidates were selected and tested for efficacy as vaccines against a challenge infecting 100% of unvaccinated ewes. Protection by R vaccines was 54% or less whereas Rev 1 afforded 100% protection. One-third of R mutant vaccinated ewes became positive in an enzyme-linked immunosorbent assay with smooth lipopolysaccharide due to the core epitopes remaining in the mutated lipopolysaccharide. We conclude that R vaccines interfere in lipopolysaccharide immunosorbent assays and are less effective than Rev 1 against B. melitensis infection of sheep.