Reactive nitrogen species (RNS) play an essential role in host defence against Mycobacterium tuberculosis (MTB) in the mouse model of tuberculosis (TB), as evidenced by the increased susceptibility of mice deficient in the inducible isoform of nitric oxide synthase (NOS2). In contrast, the role of reactive oxygen species (ROS) in protection against MTB is less clear, and mice defective in the ROS-generating phagocyte NADPH oxidase (Phox) are relatively resistant. This suggests that MTB might possess efficient mechanisms to evade or counter the phagocyte oxidative burst, effectively masking the impact of this host defence mechanism. In order to assess the role of ROS detoxification pathways in MTB virulence, we generated a katG null mutant of MTB, deficient in the KatG catalase-peroxidase-peroxynitritase, and evaluated the mutant's ability to replicate and persist in macrophages and mice. Although markedly attenuated in wild-type C57Bl/6 mice and NOS2(-/-) mice, the DeltakatG MTB strain was indistinguishable from wild-type MTB in its ability to replicate and persist in gp91(Phox-/-) mice lacking the gp91 subunit of NADPH oxidase. Similar observations were made with murine bone marrow macrophages infected ex vivo: growth of the DeltakatG MTB strain was impaired in macrophages from C57Bl/6 and NOS2(-/-) mice, but indistinguishable from wild-type MTB in gp91(Phox-/-) macrophages. These results indicate that the major role of KatG in MTB pathogenesis is to catabolize the peroxides generated by the phagocyte NADPH oxidase; in the absence of this host antimicrobial mechanism, KatG is apparently dispensable.