Novel molecular tools and genetic methods were developed to isolate genomic fragments of Mycobacterium tuberculosis that may be associated with virulence. We sought to restore virulence, a characteristic of M. tuberculosis that is correlated with growth rate in mouse spleen and lung tissue, to the avirulent strain H37Ra by complementation. A representative library of the virulent M. tuberculosis strain H37Rv was constructed and transformed into H37Ra. Enrichment for individual faster-growing recombinants was achieved by passage of pools of H37Ra transformants harboring the H37Rv library through mice. A molecular strategy was devised to isolate and clone the H37Rv genomic DNA fragment ivg, which conferred a more rapid in vivo growth rate to H37Ra.Tuberculosis is a worldwide health problem that causes approximately 3 million deaths each year (13); however, little is known about the molecular basis of the pathogenesis of tuberculosis. The disease is caused by infection with Mycobacterium tuberculosis; tubercle bacilli are inhaled and then ingested by alveolar macrophages. As is the case with most pathogens, infection with M. tuberculosis does not always result in disease. The infection is often arrested by a developing cell-mediated immunity resulting in the formation of microscopic lesions, or tubercles, in the lung. If cell-mediated immunity does not limit the spread of M. tuberculosis, caseous necrosis, bronchial wall erosion, and pulmonary cavitation may occur (5). The factors that determine whether infection with M. tuberculosis results in disease are incompletely understood.The ability to transfer and express recombinant DNA among the mycobacteria, which has been recently demonstrated (11,20), enables the usage of molecular genetics to elucidate pathogenic mechanisms. However, the present lack of evidence of homologous recombination in the pathogenic mycobacteria prevents the application of allele exchange systems (12). We sought to develop an in vivo genetic complementation system that utilizes integrating shuttle cosmid libraries to identify potential virulence genes of M. tuberculosis. One of the first examples of in vivo selection for virulent bacteria was demonstrated by the classic work of Griffith in 1928 (7). Griffith observed that as a result of genetic exchange between bacteria, virulent, capsulated pneumococci were recovered from mice infected with a mixture of live attenuated, noncap-* Corresponding author. Mailing address: § Present address: PathoGenesis Corporation, Seattle, WA 98119. sulated pneumococci and heat-killed capsulated pneumococci (7).A strategy to identify virulence determinants by genetic complementation requires (i) two strains that are genetically similar, (ii) a phenotype associated with virulence, and (iii) gene transfer systems. An existing pair of M. tuberculosis strains, H37Rv (virulent) and H37Ra (avirulent), distinguishable by their ability to cause disease in animal models (17), was used. H37Ra and H37Rv were derived from the same clinical isolate in 1934 (17,21), and puls...