Order of authorship in the original publication: Mahairas, Sabo, Hickey.... From the Abstract:

"The live attenuated bacillus Calmette-Guérin (BCG) vaccine for the prevention of disease associated with Mycobacterium tuberculosis was derived from the closely related virulent tubercle bacillus, Mycobacterium bovis. Although the BCG vaccine has been one of the most widely used vaccines in the world for over 40 years, the genetic basis of BCG's attenuation has never been elucidated. We employed subtractive genomic hybridization to identify genetic differences between virulent M. bovis and M. tuberculosis and avirulent BCG. Three distinct genomic regions of difference (designated RD1 to RD3) were found to be deleted from BCG, and the precise junctions and DNA sequence of each deletion were determined."

Mahairas and colleagues showed that the BCG bacterium had a deletion in its DNA called the RD1 deletion, which was not present in Mycobacterium tuberculosis. Because of this deletion the protein ESAT-6 could be expressed by M. tuberculosis, but not by BCG. This paved the way for a new and more accurate diagnostic test for TB based on detection of the ESAT-6 protein. This was particularly useful since it had been discovered that the Tuberculin Skin Test (TST) had a major drawback--it tested positive in individuals who had received the BCG vaccine but were not infected with TB.

(Thanks to Ron Cox for this reference and its interpretation.)